Streams/streams
Streams/streams
1
0
Fork 0
mirror of https://codeberg.org/streams/streams.git synced 2024-09-19 22:35:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
streams/install/schema_postgres.sql

1693 lines
63 KiB
MySQL
Raw Normal View History

abconfig
2016-03-01 03:31:52 +00:00
CREATE TABLE "abconfig" (
"id" serial NOT NULL,
the abconfig defaults weren't changed in the schema files when they were changed in the code
2016-07-12 20:15:20 +00:00
"chan" bigint NOT NULL DEFAULT '0',
abconfig
2016-03-01 03:31:52 +00:00
"xchan" text NOT NULL,
"cat" text NOT NULL,
"k" text NOT NULL,
"v" text NOT NULL,
PRIMARY KEY ("id")
);
create index "abconfig_chan" on abconfig ("chan");
create index "abconfig_xchan" on abconfig ("xchan");
create index "abconfig_cat" on abconfig ("cat");
create index "abconfig_k" on abconfig ("k");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "abook" (
"abook_id" serial NOT NULL,
"abook_account" bigint NOT NULL,
"abook_channel" bigint NOT NULL,
"abook_xchan" text NOT NULL DEFAULT '',
missing abook_{my,their}_perms in pg schema and missing keys in mysql schema
2017-07-05 08:36:32 +00:00
"abook_my_perms" bigint NOT NULL,
"abook_their_perms" bigint NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"abook_closeness" numeric(3) NOT NULL DEFAULT '99',
"abook_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"abook_updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"abook_connected" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"abook_dob" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"abook_flags" bigint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"abook_blocked" smallint NOT NULL DEFAULT '0',
"abook_ignored" smallint NOT NULL DEFAULT '0',
"abook_hidden" smallint NOT NULL DEFAULT '0',
"abook_archived" smallint NOT NULL DEFAULT '0',
"abook_pending" smallint NOT NULL DEFAULT '0',
"abook_unconnected" smallint NOT NULL DEFAULT '0',
"abook_self" smallint NOT NULL DEFAULT '0',
as fixes
2018-06-18 02:25:45 +00:00
"abook_rself" smallint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"abook_feed" smallint NOT NULL DEFAULT '0',
abook_not_here flag created to indicate singleton connections which are connected to this channel but not on this hub. abook_instance enumerates which hubs the connections is valid, but we ultimately need something more efficiently searchable to decide what operations are supported w/r/t this connection in the context of this hub. This flag is ignored during sync to clones although the code to set it correctly during channel creation, import, and sync has not yet been implemented.
2017-05-04 01:50:32 +00:00
"abook_not_here" smallint NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"abook_profile" varchar(64) NOT NULL DEFAULT '',
no this isn't it. well ok, it's part of it, but not the important part. it's still pretty cool.
2015-06-29 04:16:56 +00:00
"abook_incl" TEXT NOT NULL DEFAULT '',
"abook_excl" TEXT NOT NULL DEFAULT '',
move the instance stuff to abook where it belongs
2015-12-05 21:24:58 +00:00
"abook_instance" TEXT NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("abook_id")
);
create index "abook_account" on abook ("abook_account");
create index "abook_channel" on abook ("abook_channel");
create index "abook_xchan" on abook ("abook_xchan");
missing abook_{my,their}_perms in pg schema and missing keys in mysql schema
2017-07-05 08:36:32 +00:00
create index "abook_my_perms" on abook ("abook_my_perms");
create index "abook_their_perms" on abook ("abook_their_perms");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "abook_closeness" on abook ("abook_closeness");
create index "abook_created" on abook ("abook_created");
create index "abook_updated" on abook ("abook_updated");
create index "abook_flags" on abook ("abook_flags");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "abook_blocked" on abook ("abook_blocked");
create index "abook_ignored" on abook ("abook_ignored");
create index "abook_hidden" on abook ("abook_hidden");
create index "abook_archived" on abook ("abook_archived");
create index "abook_pending" on abook ("abook_pending");
create index "abook_unconnected" on abook ("abook_unconnected");
create index "abook_self" on abook ("abook_self");
create index "abook_feed" on abook ("abook_feed");
abook_not_here flag created to indicate singleton connections which are connected to this channel but not on this hub. abook_instance enumerates which hubs the connections is valid, but we ultimately need something more efficiently searchable to decide what operations are supported w/r/t this connection in the context of this hub. This flag is ignored during sync to clones although the code to set it correctly during channel creation, import, and sync has not yet been implemented.
2017-05-04 01:50:32 +00:00
create index "abook_not_here" on abook ("abook_not_here");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "abook_profile" on abook ("abook_profile");
create index "abook_dob" on abook ("abook_dob");
create index "abook_connected" on abook ("abook_connected");
sql optimisation for affinity searches. A new index was added which wasn't added retro-actively to existing DBs as an update. It isn't clear if this helps sites any more than just restricting the abook table to certain channel_id's is (and this field is already indexed).
2015-03-11 01:55:56 +00:00
create index "abook_channel_closeness" on abook ("abook_channel", "abook_closeness");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "account" (
"account_id" serial NOT NULL,
"account_parent" bigint NOT NULL DEFAULT '0',
"account_default_channel" bigint NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"account_salt" varchar(32) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"account_password" text NOT NULL DEFAULT '',
"account_email" text NOT NULL DEFAULT '',
"account_external" text NOT NULL DEFAULT '',
"account_language" varchar(16) NOT NULL DEFAULT 'en',
"account_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"account_lastlog" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"account_flags" bigint NOT NULL DEFAULT '0',
"account_roles" bigint NOT NULL DEFAULT '0',
"account_reset" text NOT NULL DEFAULT '',
"account_expires" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"account_expire_notified" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"account_service_class" varchar(32) NOT NULL DEFAULT '',
"account_level" bigint NOT NULL DEFAULT '0',
"account_password_changed" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("account_id")
);
create index "account_email" on account ("account_email");
create index "account_service_class" on account ("account_service_class");
create index "account_parent" on account ("account_parent");
create index "account_flags" on account ("account_flags");
create index "account_roles" on account ("account_roles");
create index "account_lastlog" on account ("account_lastlog");
create index "account_expires" on account ("account_expires");
create index "account_default_channel" on account ("account_default_channel");
create index "account_external" on account ("account_external");
create index "account_level" on account ("account_level");
create index "account_password_changed" on account ("account_password_changed");
CREATE TABLE "addon" (
"id" serial NOT NULL,
start removing reserved words from database column names (this run: addon and hook)
2016-05-31 02:44:30 +00:00
"aname" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"version" text NOT NULL DEFAULT '0',
"installed" numeric(1) NOT NULL DEFAULT '0',
"hidden" numeric(1) NOT NULL DEFAULT '0',
start removing reserved words from database column names (this run: addon and hook)
2016-05-31 02:44:30 +00:00
"tstamp" numeric(20) NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"plugin_admin" numeric(1) NOT NULL DEFAULT '0',
PRIMARY KEY ("id")
);
create index "addon_hidden_idx" on addon ("hidden");
start removing reserved words from database column names (this run: addon and hook)
2016-05-31 02:44:30 +00:00
create index "addon_name_idx" on addon ("aname");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "addon_installed_idx" on addon ("installed");
CREATE TABLE "app" (
"id" serial NOT NULL,
"app_id" text NOT NULL DEFAULT '',
"app_sig" text NOT NULL DEFAULT '',
"app_author" text NOT NULL DEFAULT '',
"app_name" text NOT NULL DEFAULT '',
issue #446 apps usability - disable app if attached to a plugin that is uninstalled, allow system apps to be soft deleted and undeleted from the edit pane.
2016-12-15 03:11:01 +00:00
"app_desc" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"app_url" text NOT NULL DEFAULT '',
"app_photo" text NOT NULL DEFAULT '',
"app_version" text NOT NULL DEFAULT '',
"app_channel" bigint NOT NULL DEFAULT '0',
"app_addr" text NOT NULL DEFAULT '',
"app_price" text NOT NULL DEFAULT '',
"app_page" text NOT NULL DEFAULT '',
"app_requires" text NOT NULL DEFAULT '',
some preliminary structural work for app organisation
2016-05-04 01:41:16 +00:00
"app_deleted" smallint NOT NULL DEFAULT '0',
"app_system" smallint NOT NULL DEFAULT '0',
issue #446 apps usability - disable app if attached to a plugin that is uninstalled, allow system apps to be soft deleted and undeleted from the edit pane.
2016-12-15 03:11:01 +00:00
"app_plugin" text NOT NULL DEFAULT '',
add app_options field
2018-08-03 05:49:30 +00:00
"app_options" smallint NOT NULL DEFAULT '0',
add timestamps for syncing apps
2015-09-03 04:07:24 +00:00
"app_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"app_edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
create index "app_id" on app ("app_id");
create index "app_name" on app ("app_name");
create index "app_url" on app ("app_url");
create index "app_photo" on app ("app_photo");
create index "app_version" on app ("app_version");
create index "app_channel" on app ("app_channel");
create index "app_price" on app ("app_price");
add timestamps for syncing apps
2015-09-03 04:07:24 +00:00
create index "app_created" on app ("app_created");
create index "app_edited" on app ("app_edited");
some preliminary structural work for app organisation
2016-05-04 01:41:16 +00:00
create index "app_deleted" on app ("app_deleted");
create index "app_system" on app ("app_system");
schema updates for zot access tokens
2016-07-15 03:51:15 +00:00
issue #446 apps usability - disable app if attached to a plugin that is uninstalled, allow system apps to be soft deleted and undeleted from the edit pane.
2016-12-15 03:11:01 +00:00
schema updates for zot access tokens
2016-07-15 03:51:15 +00:00
CREATE TABLE "atoken" (
"atoken_id" serial NOT NULL,
"atoken_aid" bigint NOT NULL DEFAULT 0,
"atoken_uid" bigint NOT NULL DEFAULT 0,
"atoken_name" varchar(255) NOT NULL DEFAULT '',
"atoken_token" varchar(255) NOT NULL DEFAULT '',
"atoken_expires" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("atoken_id"));
create index atoken_aid on atoken (atoken_aid);
create index atoken_uid on atoken (atoken_uid);
create index atoken_name on atoken (atoken_name);
create index atoken_token on atoken (atoken_token);
create index atoken_expires on atoken (atoken_expires);
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "attach" (
"id" serial NOT NULL,
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL DEFAULT '0',
"hash" varchar(64) NOT NULL DEFAULT '',
"creator" varchar(128) NOT NULL DEFAULT '',
"filename" text NOT NULL DEFAULT '',
"filetype" varchar(64) NOT NULL DEFAULT '',
"filesize" bigint NOT NULL DEFAULT '0',
"revision" bigint NOT NULL DEFAULT '0',
"folder" varchar(64) NOT NULL DEFAULT '',
"flags" bigint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"is_dir" smallint NOT NULL DEFAULT '0',
"is_photo" smallint NOT NULL DEFAULT '0',
"os_storage" smallint NOT NULL DEFAULT '0',
"os_path" text NOT NULL,
typo
2017-04-13 01:48:06 +00:00
"display_path" text NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"content" bytea NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
PRIMARY KEY ("id")
);
create index "attach_aid_idx" on attach ("aid");
create index "attach_uid_idx" on attach ("uid");
create index "attach_hash_idx" on attach ("hash");
create index "attach_filename_idx" on attach ("filename");
create index "attach_filetype_idx" on attach ("filetype");
create index "attach_filesize_idx" on attach ("filesize");
create index "attach_created_idx" on attach ("created");
create index "attach_edited_idx" on attach ("edited");
create index "attach_revision_idx" on attach ("revision");
create index "attach_folder_idx" on attach ("folder");
create index "attach_flags_idx" on attach ("flags");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "attach_is_dir_idx" on attach ("is_dir");
create index "attach_is_photo_idx" on attach ("is_photo");
create index "attach_os_storage_idx" on attach ("os_storage");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "attach_creator_idx" on attach ("creator");
CREATE TABLE "auth_codes" (
"id" varchar(40) NOT NULL,
"client_id" varchar(20) NOT NULL,
"redirect_uri" varchar(200) NOT NULL,
"expires" bigint NOT NULL,
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"auth_scope" varchar(512) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
CREATE TABLE "cache" (
"k" text NOT NULL,
"v" text NOT NULL,
"updated" timestamp NOT NULL,
PRIMARY KEY ("k")
);
schema changes needed for the caldav client bit
2016-06-18 22:33:47 +00:00
CREATE TABLE "cal" (
"cal_id" serial NOT NULL,
"cal_aid" bigint NOT NULL DEFAULT '0',
"cal_uid" bigint NOT NULL DEFAULT '0',
"cal_hash" text NOT NULL,
"cal_name" text NOT NULL,
"uri" text NOT NULL,
"logname" text NOT NULL,
"pass" text NOT NULL,
"ctag" text NOT NULL,
"synctoken" text NOT NULL,
issue #511, postgres schema issues
2016-09-10 04:25:57 +00:00
"cal_types" text NOT NULL DEFAULT '0',
schema changes needed for the caldav client bit
2016-06-18 22:33:47 +00:00
PRIMARY KEY ("cal_id")
);
create index "cal_hash_idx" on cal ("cal_hash");
create index "cal_name_idx" on cal ("cal_name");
create index "cal_types_idx" on cal ("cal_types");
create index "cal_aid_idx" on cal ("cal_aid");
create index "cal_uid_idx" on cal ("cal_uid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "channel" (
"channel_id" serial NOT NULL,
"channel_account_id" bigint NOT NULL DEFAULT '0',
"channel_primary" numeric(1) NOT NULL DEFAULT '0',
"channel_name" text NOT NULL DEFAULT '',
"channel_address" text NOT NULL DEFAULT '',
"channel_guid" text NOT NULL DEFAULT '',
"channel_guid_sig" text NOT NULL,
"channel_hash" text NOT NULL DEFAULT '',
"channel_timezone" varchar(128) NOT NULL DEFAULT 'UTC',
"channel_location" text NOT NULL DEFAULT '',
"channel_theme" text NOT NULL DEFAULT '',
"channel_startpage" text NOT NULL DEFAULT '',
"channel_pubkey" text NOT NULL,
"channel_prvkey" text NOT NULL,
"channel_notifyflags" bigint NOT NULL DEFAULT '65535',
"channel_pageflags" bigint NOT NULL DEFAULT '0',
"channel_dirdate" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
add channel_lastpost timestamp to help optimise some outrageously expensive queries.
2015-03-27 02:55:52 +00:00
"channel_lastpost" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"channel_deleted" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
Active channels information is a bit imprecise. Provide a higher accuracy method. This will require a transition period
2018-04-17 04:40:43 +00:00
"channel_active" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"channel_max_anon_mail" bigint NOT NULL DEFAULT '10',
"channel_max_friend_req" bigint NOT NULL DEFAULT '10',
"channel_expire_days" bigint NOT NULL DEFAULT '0',
"channel_passwd_reset" text NOT NULL DEFAULT '',
"channel_default_group" text NOT NULL DEFAULT '',
"channel_allow_cid" text ,
"channel_allow_gid" text ,
"channel_deny_cid" text ,
"channel_deny_gid" text ,
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"channel_removed" smallint NOT NULL DEFAULT '0',
"channel_system" smallint NOT NULL DEFAULT '0',
schema change to support channel move (a completely different operation than channel clone)
2016-02-25 00:59:50 +00:00
"channel_moved" text NOT NULL DEFAULT '',
backend infrastructure for 'channel protection password'; which will be used to optionally encrypt export files and resolve channel/identity ownership/hijacking disputes
2017-04-13 00:32:28 +00:00
"channel_password" varchar(255) NOT NULL,
"channel_salt" varchar(255) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("channel_id"),
UNIQUE ("channel_address")
);
create index "channel_account_id" on channel ("channel_account_id");
create index "channel_primary" on channel ("channel_primary");
create index "channel_name" on channel ("channel_name");
create index "channel_timezone" on channel ("channel_timezone");
create index "channel_location" on channel ("channel_location");
create index "channel_theme" on channel ("channel_theme");
create index "channel_notifyflags" on channel ("channel_notifyflags");
create index "channel_pageflags" on channel ("channel_pageflags");
create index "channel_max_anon_mail" on channel ("channel_max_anon_mail");
create index "channel_max_friend_req" on channel ("channel_max_friend_req");
create index "channel_default_gid" on channel ("channel_default_group");
create index "channel_guid" on channel ("channel_guid");
create index "channel_hash" on channel ("channel_hash");
create index "channel_expire_days" on channel ("channel_expire_days");
create index "channel_deleted" on channel ("channel_deleted");
Active channels information is a bit imprecise. Provide a higher accuracy method. This will require a transition period
2018-04-17 04:40:43 +00:00
create index "channel_active" on channel ("channel_active");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "channel_dirdate" on channel ("channel_dirdate");
add channel_lastpost timestamp to help optimise some outrageously expensive queries.
2015-03-27 02:55:52 +00:00
create index "channel_lastpost" on channel ("channel_lastpost");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "channel_removed" on channel ("channel_removed");
create index "channel_system" on channel ("channel_system");
schema change to support channel move (a completely different operation than channel clone)
2016-02-25 00:59:50 +00:00
create index "channel_moved" on channel ("channel_moved");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "chat" (
"chat_id" serial NOT NULL,
"chat_room" bigint NOT NULL DEFAULT '0',
"chat_xchan" text NOT NULL DEFAULT '',
"chat_text" text NOT NULL,
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("chat_id")
);
create index "chat_room_idx" on chat ("chat_room");
create index "chat_xchan_idx" on chat ("chat_xchan");
create index "chat_created_idx" on chat ("created");
CREATE TABLE "chatpresence" (
"cp_id" serial NOT NULL,
"cp_room" bigint NOT NULL DEFAULT '0',
"cp_xchan" text NOT NULL DEFAULT '',
"cp_last" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"cp_status" text NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"cp_client" varchar(128) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("cp_id")
);
create index "cp_room" on chatpresence ("cp_room");
create index "cp_xchan" on chatpresence ("cp_xchan");
create index "cp_last" on chatpresence ("cp_last");
create index "cp_status" on chatpresence ("cp_status");
CREATE TABLE "chatroom" (
"cr_id" serial NOT NULL,
"cr_aid" bigint NOT NULL DEFAULT '0',
"cr_uid" bigint NOT NULL DEFAULT '0',
"cr_name" text NOT NULL DEFAULT '',
"cr_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"cr_edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"cr_expire" bigint NOT NULL DEFAULT '0',
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
PRIMARY KEY ("cr_id")
);
create index "cr_aid" on chatroom ("cr_aid");
create index "cr_uid" on chatroom ("cr_uid");
create index "cr_name" on chatroom ("cr_name");
create index "cr_created" on chatroom ("cr_created");
create index "cr_edited" on chatroom ("cr_edited");
create index "cr_expire" on chatroom ("cr_expire");
CREATE TABLE "clients" (
"client_id" varchar(20) NOT NULL,
"pw" varchar(20) NOT NULL,
"redirect_uri" varchar(200) NOT NULL,
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"clname" text,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"icon" text,
"uid" bigint NOT NULL DEFAULT '0',
PRIMARY KEY ("client_id")
);
CREATE TABLE "config" (
"id" serial NOT NULL,
"cat" text NOT NULL,
"k" text NOT NULL,
"v" text NOT NULL,
PRIMARY KEY ("id"),
UNIQUE ("cat","k")
);
CREATE TABLE "conv" (
"id" serial NOT NULL,
"guid" text NOT NULL,
"recips" text NOT NULL,
"uid" bigint NOT NULL,
"creator" text NOT NULL,
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"subject" text NOT NULL,
PRIMARY KEY ("id")
);
create index "conv_created_idx" on conv ("created");
create index "conv_updated_idx" on conv ("updated");
DB table for delivery reports
2015-09-19 05:55:47 +00:00
CREATE TABLE IF NOT EXISTS "dreport" (
issue #523
2016-09-23 00:00:32 +00:00
"dreport_id" serial NOT NULL,
updates to Postgres schema
2015-11-11 15:42:41 +00:00
"dreport_channel" int NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"dreport_mid" varchar(255) NOT NULL DEFAULT '',
"dreport_site" varchar(255) NOT NULL DEFAULT '',
"dreport_recip" varchar(255) NOT NULL DEFAULT '',
dreport updates - separate dreport_name from dreport_recip
2018-06-25 02:41:03 +00:00
"dreport_name" varchar(255) NOT NULL DEFAULT '',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"dreport_result" varchar(255) NOT NULL DEFAULT '',
DB table for delivery reports
2015-09-19 05:55:47 +00:00
"dreport_time" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"dreport_xchan" varchar(255) NOT NULL DEFAULT '',
"dreport_queue" varchar(255) NOT NULL DEFAULT '',
DB table for delivery reports
2015-09-19 05:55:47 +00:00
PRIMARY KEY ("dreport_id")
);
create index "dreport_mid" on dreport ("dreport_mid");
create index "dreport_site" on dreport ("dreport_site");
create index "dreport_time" on dreport ("dreport_time");
create index "dreport_xchan" on dreport ("dreport_xchan");
heavy lifting for delivery reports
2015-09-22 09:32:04 +00:00
create index "dreport_queue" on dreport ("dreport_queue");
DB table for delivery reports
2015-09-19 05:55:47 +00:00
create index "dreport_channel" on dreport ("dreport_channel");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "event" (
"id" serial NOT NULL,
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL,
schema changes needed for the caldav client bit
2016-06-18 22:33:47 +00:00
"cal_id" bigint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"event_xchan" text NOT NULL DEFAULT '',
"event_hash" text NOT NULL DEFAULT '',
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"dtstart" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"dtend" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"summary" text NOT NULL,
"description" text NOT NULL,
"location" text NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"etype" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"nofinish" numeric(1) NOT NULL DEFAULT '0',
"adjust" numeric(1) NOT NULL DEFAULT '1',
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"dismissed" numeric(1) NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"event_status" varchar(255) NOT NULL DEFAULT '',
add DB support for tasks, todo items and repeating events
2015-08-17 00:17:34 +00:00
"event_status_date" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
updates to Postgres schema
2015-11-11 15:42:41 +00:00
"event_percent" smallint NOT NULL DEFAULT '0',
add DB support for tasks, todo items and repeating events
2015-08-17 00:17:34 +00:00
"event_repeat" text NOT NULL,
add event_sequence for vtodo's
2015-08-19 00:10:40 +00:00
"event_sequence" smallint NOT NULL DEFAULT '0',
missed this one - it's required
2015-08-21 05:13:25 +00:00
"event_priority" smallint NOT NULL DEFAULT '0',
add vdata field for caldav
2015-09-24 05:32:46 +00:00
"event_vdata" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
create index "event_uid_idx" on event ("uid");
schema changes needed for the caldav client bit
2016-06-18 22:33:47 +00:00
create index "event_cal_idx" on event ("cal_id");
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "event_etype_idx" on event ("etype");
create index "event_dtstart_idx" on event ("dtstart");
create index "event_dtend_idx" on event ("dtend");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "event_adjust_idx" on event ("adjust");
create index "event_nofinish_idx" on event ("nofinish");
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "event_dismissed_idx" on event ("dismissed");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "event_aid_idx" on event ("aid");
create index "event_hash_idx" on event ("event_hash");
create index "event_xchan_idx" on event ("event_xchan");
add DB support for tasks, todo items and repeating events
2015-08-17 00:17:34 +00:00
create index "event_status_idx" on event ("event_status");
add event_sequence for vtodo's
2015-08-19 00:10:40 +00:00
create index "event_sequence_idx" on event ("event_sequence");
missed this one - it's required
2015-08-21 05:13:25 +00:00
create index "event_priority_idx" on event ("event_priority");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
rename groups and group_members tables
2018-09-26 00:47:43 +00:00
CREATE TABLE "pgrp_member" (
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"id" serial NOT NULL,
"uid" bigint NOT NULL,
"gid" bigint NOT NULL,
"xchan" text NOT NULL DEFAULT '',
PRIMARY KEY ("id")
);
postgres pgrp fix
2018-09-28 00:13:58 +00:00
create index "groupmember_uid" on pgrp_member ("uid");
create index "groupmember_gid" on pgrp_member ("gid");
create index "groupmember_xchan" on pgrp_member ("xchan");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
rename groups and group_members tables
2018-09-26 00:47:43 +00:00
CREATE TABLE "pgrp" (
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"id" serial NOT NULL,
"hash" text NOT NULL DEFAULT '',
"uid" bigint NOT NULL,
"visible" numeric(1) NOT NULL DEFAULT '0',
"deleted" numeric(1) NOT NULL DEFAULT '0',
more db column renames
2016-06-01 04:45:33 +00:00
"gname" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
postgres pgrp fix
2018-09-28 00:13:58 +00:00
create index "groups_uid_idx" on pgrp ("uid");
create index "groups_visible_idx" on pgrp ("visible");
create index "groups_deleted_idx" on pgrp ("deleted");
create index "groups_hash_idx" on pgrp ("hash");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "hook" (
"id" serial NOT NULL,
"hook" text NOT NULL,
"file" text NOT NULL,
start removing reserved words from database column names (this run: addon and hook)
2016-05-31 02:44:30 +00:00
"fn" text NOT NULL,
allow hooks to have negative priority
2016-10-13 03:04:19 +00:00
"priority" smallint NOT NULL DEFAULT '0',
new hook interface (the old one still works but requires handlers to have two calling arguments; the first of which is no longer used). The new interface is called from Zotlabs\Extend\Hook::register() and allows you to specify which hook version to use. The default will be the new interface with one function argument. We also implement the hook priority field which was always there but needed to be set manually in the DB. This provides a way for two hook handlers that implement the same hook interface to determine which order to be called in the event of conflicts.
2016-04-22 00:03:05 +00:00
"hook_version" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
create index "hook_idx" on hook ("hook");
new hook interface (the old one still works but requires handlers to have two calling arguments; the first of which is no longer used). The new interface is called from Zotlabs\Extend\Hook::register() and allows you to specify which hook version to use. The default will be the new interface with one function argument. We also implement the hook priority field which was always there but needed to be set manually in the DB. This provides a way for two hook handlers that implement the same hook interface to determine which order to be called in the event of conflicts.
2016-04-22 00:03:05 +00:00
create index "hook_version_idx" on hook ("hook_version");
allow hooks to have negative priority
2016-10-13 03:04:19 +00:00
create index "hook_priority_idx" on hook ("priority");
hubloc DB changes needed for z6
2018-08-14 04:18:20 +00:00
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "hubloc" (
"hubloc_id" serial NOT NULL,
"hubloc_guid" text NOT NULL DEFAULT '',
"hubloc_guid_sig" text NOT NULL DEFAULT '',
more zot6 work
2018-05-24 04:48:19 +00:00
"hubloc_id_url" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"hubloc_hash" text NOT NULL,
"hubloc_addr" text NOT NULL DEFAULT '',
"hubloc_network" text NOT NULL DEFAULT '',
"hubloc_flags" bigint NOT NULL DEFAULT '0',
"hubloc_status" bigint NOT NULL DEFAULT '0',
"hubloc_url" text NOT NULL DEFAULT '',
"hubloc_url_sig" text NOT NULL DEFAULT '',
start fresh after a lot of heavy lifting and fix a lot of stuff
2018-06-04 04:23:34 +00:00
"hubloc_site_id" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"hubloc_host" text NOT NULL DEFAULT '',
"hubloc_callback" text NOT NULL DEFAULT '',
"hubloc_connect" text NOT NULL DEFAULT '',
"hubloc_sitekey" text NOT NULL DEFAULT '',
"hubloc_updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"hubloc_connected" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"hubloc_primary" smallint NOT NULL DEFAULT '0',
"hubloc_orphancheck" smallint NOT NULL DEFAULT '0',
"hubloc_error" smallint NOT NULL DEFAULT '0',
"hubloc_deleted" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("hubloc_id")
);
create index "hubloc_url" on hubloc ("hubloc_url");
start fresh after a lot of heavy lifting and fix a lot of stuff
2018-06-04 04:23:34 +00:00
create index "hubloc_site_id" on hubloc ("hubloc_site_id");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "hubloc_guid" on hubloc ("hubloc_guid");
more zot6 work
2018-05-24 04:48:19 +00:00
create index "hubloc_id_url" on hubloc ("hubloc_id_url");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "hubloc_flags" on hubloc ("hubloc_flags");
create index "hubloc_connect" on hubloc ("hubloc_connect");
create index "hubloc_host" on hubloc ("hubloc_host");
create index "hubloc_addr" on hubloc ("hubloc_addr");
create index "hubloc_network" on hubloc ("hubloc_network");
create index "hubloc_updated" on hubloc ("hubloc_updated");
create index "hubloc_connected" on hubloc ("hubloc_connected");
create index "hubloc_status" on hubloc ("hubloc_status");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "hubloc_primary" on hubloc ("hubloc_primary");
create index "hubloc_orphancheck" on hubloc ("hubloc_orphancheck");
create index "hubloc_error" on hubloc ("hubloc_error");
create index "hubloc_deleted" on hubloc ("hubloc_deleted");
schema updates for iconfig
2016-02-18 00:24:26 +00:00
CREATE TABLE "iconfig" (
"id" serial NOT NULL,
"iid" bigint NOT NULL DEFAULT '0',
"cat" text NOT NULL DEFAULT '',
"k" text NOT NULL DEFAULT '',
"v" text NOT NULL DEFAULT '',
iconfig - add sharing variable
2016-02-18 08:20:08 +00:00
"sharing" int NOT NULL DEFAULT '0',
fix typo in postgres schema
2016-03-01 16:54:52 +00:00
PRIMARY KEY("id")
schema updates for iconfig
2016-02-18 00:24:26 +00:00
);
create index "iconfig_iid" on iconfig ("iid");
create index "iconfig_cat" on iconfig ("cat");
create index "iconfig_k" on iconfig ("k");
iconfig - add sharing variable
2016-02-18 08:20:08 +00:00
create index "iconfig_sharing" on iconfig ("sharing");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "issue" (
"issue_id" serial NOT NULL,
"issue_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"issue_updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"issue_assigned" text NOT NULL,
"issue_priority" bigint NOT NULL,
"issue_status" bigint NOT NULL,
"issue_component" text NOT NULL,
PRIMARY KEY ("issue_id")
);
create index "issue_created" on issue ("issue_created");
create index "issue_updated" on issue ("issue_updated");
create index "issue_assigned" on issue ("issue_assigned");
create index "issue_priority" on issue ("issue_priority");
create index "issue_status" on issue ("issue_status");
create index "issue_component" on issue ("issue_component");
CREATE TABLE "item" (
"id" serial NOT NULL,
"mid" text NOT NULL DEFAULT '',
add item uuid
2019-02-19 22:59:22 +00:00
"uuid" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL DEFAULT '0',
"parent" bigint NOT NULL DEFAULT '0',
"parent_mid" text NOT NULL DEFAULT '',
"thr_parent" text NOT NULL DEFAULT '',
store item_level to avoid a lot of recursion during rendering
2019-01-14 22:47:05 +00:00
"item_level" bigint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"expires" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"commented" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"received" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"changed" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"comments_closed" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"owner_xchan" text NOT NULL DEFAULT '',
"author_xchan" text NOT NULL DEFAULT '',
"source_xchan" text NOT NULL DEFAULT '',
"mimetype" text NOT NULL DEFAULT '',
"title" text NOT NULL,
add summary field to item table
2018-07-16 01:31:39 +00:00
"summary" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"body" text NOT NULL,
convert ITEM_WALL from bitfield to standalone
2015-05-07 04:03:33 +00:00
"html" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"app" text NOT NULL DEFAULT '',
"lang" varchar(64) NOT NULL DEFAULT '',
"revision" bigint NOT NULL DEFAULT '0',
"verb" text NOT NULL DEFAULT '',
"obj_type" text NOT NULL DEFAULT '',
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"obj" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"tgt_type" text NOT NULL DEFAULT '',
"target" text NOT NULL,
"layout_mid" text NOT NULL DEFAULT '',
"postopts" text NOT NULL DEFAULT '',
"route" text NOT NULL DEFAULT '',
"llink" text NOT NULL DEFAULT '',
"plink" text NOT NULL DEFAULT '',
"resource_id" text NOT NULL DEFAULT '',
"resource_type" varchar(16) NOT NULL DEFAULT '',
"attach" text NOT NULL,
"sig" text NOT NULL DEFAULT '',
"location" text NOT NULL DEFAULT '',
"coord" text NOT NULL DEFAULT '',
"public_policy" text NOT NULL DEFAULT '',
"comment_policy" text NOT NULL DEFAULT '',
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
"item_restrict" bigint NOT NULL DEFAULT '0',
"item_flags" bigint NOT NULL DEFAULT '0',
"item_private" numeric(4) NOT NULL DEFAULT '0',
fix issues in schema_postgres.sql - change PRIMARY_KEY to PRIMARY KEY - replace smallint(1) with smallint
2015-02-26 22:50:46 +00:00
"item_unseen" smallint NOT NULL DEFAULT '0',
convert ITEM_WALL from bitfield to standalone
2015-05-07 04:03:33 +00:00
"item_wall" smallint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"item_origin" smallint NOT NULL DEFAULT '0',
"item_starred" smallint NOT NULL DEFAULT '0',
"item_uplink" smallint NOT NULL DEFAULT '0',
"item_consensus" smallint NOT NULL DEFAULT '0',
"item_thread_top" smallint NOT NULL DEFAULT '0',
"item_notshown" smallint NOT NULL DEFAULT '0',
"item_nsfw" smallint NOT NULL DEFAULT '0',
"item_relay" smallint NOT NULL DEFAULT '0',
"item_mentionsme" smallint NOT NULL DEFAULT '0',
"item_nocomment" smallint NOT NULL DEFAULT '0',
"item_obscured" smallint NOT NULL DEFAULT '0',
"item_verified" smallint NOT NULL DEFAULT '0',
"item_retained" smallint NOT NULL DEFAULT '0',
"item_rss" smallint NOT NULL DEFAULT '0',
"item_deleted" smallint NOT NULL DEFAULT '0',
updates to Postgres schema
2015-11-11 15:42:41 +00:00
"item_type" int NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"item_hidden" smallint NOT NULL DEFAULT '0',
"item_unpublished" smallint NOT NULL DEFAULT '0',
"item_delayed" smallint NOT NULL DEFAULT '0',
"item_pending_remove" smallint NOT NULL DEFAULT '0',
"item_blocked" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"item_search_vector" tsvector,
PRIMARY KEY ("id")
);
create index "item_uid" on item ("uid");
create index "item_parent" on item ("parent");
create index "item_created" on item ("created");
create index "item_edited" on item ("edited");
create index "item_received" on item ("received");
create index "item_uid_commented" on item ("uid","commented");
create index "item_uid_created" on item ("uid","created");
add combined index for item.uid and item.item_unseen. this speeds up notifications by a magnitude.
2017-09-01 19:27:59 +00:00
create index "item_uid_unseen" on item ("uid","item_unseen");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "item_changed" on item ("changed");
create index "item_comments_closed" on item ("comments_closed");
create index "item_aid" on item ("aid");
create index "item_owner_xchan" on item ("owner_xchan");
create index "item_author_xchan" on item ("author_xchan");
db update to add index to item.resource_id we query this in wikis a lot
2017-09-16 08:11:59 +00:00
create index "item_resource_id" on item ("resource_id");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "item_resource_type" on item ("resource_type");
create index "item_restrict" on item ("item_restrict");
create index "item_flags" on item ("item_flags");
create index "item_commented" on item ("commented");
create index "item_verb" on item ("verb");
db update to add index for item.obj_type
2017-07-31 14:49:56 +00:00
create index "item_obj_type" on item ("obj_type");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "item_private" on item ("item_private");
create index "item_llink" on item ("llink");
create index "item_expires" on item ("expires");
create index "item_revision" on item ("revision");
create index "item_mimetype" on item ("mimetype");
create index "item_mid" on item ("mid");
add item uuid
2019-02-19 22:59:22 +00:00
create index "item_uuid" on item ("uuid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "item_parent_mid" on item ("parent_mid");
create index "item_uid_mid" on item ("mid","uid");
create index "item_public_policy" on item ("public_policy");
create index "item_comment_policy" on item ("comment_policy");
create index "item_layout_mid" on item ("layout_mid");
provide relief to sites that are severely impacted by the slow ITEM_UNSEEN searches. This does not incorporate any other flag optimisations as that will require a major DB update and possibly involve significant downtime. This is just to bite off a little chunk now and provide some much needed relief.
2015-02-13 01:45:25 +00:00
create index "item_unseen" on item ("item_unseen");
convert ITEM_WALL from bitfield to standalone
2015-05-07 04:03:33 +00:00
create index "item_wall" on item ("item_wall");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "item_origin" on item ("item_origin");
create index "item_starred" on item ("item_starred");
create index "item_uplink" on item ("item_uplink");
create index "item_consensus" on item ("item_consensus");
create index "item_thread_top" on item ("item_thread_top");
create index "item_notshown" on item ("item_notshown");
create index "item_nsfw" on item ("item_nsfw");
create index "item_relay" on item ("item_relay");
create index "item_mentionsme" on item ("item_mentionsme");
create index "item_nocomment" on item ("item_nocomment");
create index "item_obscured" on item ("item_obscured");
create index "item_verified" on item ("item_verified");
create index "item_retained" on item ("item_retained");
create index "item_rss" on item ("item_rss");
create index "item_deleted" on item ("item_deleted");
create index "item_type" on item ("item_type");
create index "item_hidden" on item ("item_hidden");
create index "item_unpublished" on item ("item_unpublished");
create index "item_delayed" on item ("item_delayed");
create index "item_pending_remove" on item ("item_pending_remove");
create index "item_blocked" on item ("item_blocked");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
-- fulltext indexes
create index "item_search_idx" on item USING gist("item_search_vector");
create index "item_allow_cid" on item ("allow_cid");
create index "item_allow_gid" on item ("allow_gid");
create index "item_deny_cid" on item ("deny_cid");
create index "item_deny_gid" on item ("deny_gid");
CREATE TABLE "item_id" (
"id" serial NOT NULL,
"iid" bigint NOT NULL,
"uid" bigint NOT NULL,
"sid" text NOT NULL,
"service" text NOT NULL,
PRIMARY KEY ("id")
);
create index "itemid_uid" on item_id ("uid");
create index "itemid_sid" on item_id ("sid");
create index "itemid_service" on item_id ("service");
create index "itemid_iid" on item_id ("iid");
CREATE TABLE "likes" (
"id" serial NOT NULL,
"channel_id" bigint NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"liker" varchar(128) NOT NULL DEFAULT '',
"likee" varchar(128) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"iid" bigint NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"i_mid" varchar(255) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"verb" text NOT NULL DEFAULT '',
"target_type" text NOT NULL DEFAULT '',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"target_id" varchar(128) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"target" text NOT NULL,
PRIMARY KEY ("id")
);
create index "likes_channel_id" on likes ("channel_id");
create index "likes_liker" on likes ("liker");
create index "likes_likee" on likes ("likee");
create index "likes_iid" on likes ("iid");
change the likes db structure to make it more easily importable/exportable
2015-09-01 03:55:25 +00:00
create index "likes_i_mid" on likes ("i_mid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "likes_verb" on likes ("verb");
create index "likes_target_type" on likes ("target_type");
create index "likes_target_id" on likes ("target_id");
add linkid structure
2018-07-05 01:46:28 +00:00
CREATE TABLE "linkid" (
"link_id" serial NOT NULL,
"ident" text NOT NULL DEFAULT '',
"link" text NOT NULL DEFAULT '',
"ikey" text NOT NULL,
"lkey" text NOT NULL,
"isig" text NOT NULL,
"lsig" text NOT NULL,
"sigtype" smallint NOT NULL DEFAULT '0',
PRIMARY KEY ("link_id")
);
create index "linkid_ident" on linkid ("ident");
create index "linkid_link" on linkid ("link");
important hyperdrive component
2018-09-10 03:53:20 +00:00
CREATE TABLE listeners (
id serial NOT NULL,
target_id text NOT NULL,
portable_id text NOT NULL,
ltype smallint NOT NULL DEFAULT '0',
PRIMARY KEY (id)
);
create index "target_id_idx" on listeners ("target_id");
create index "portable_id_idx" on listeners ("portable_id");
create index "ltype_idx" on listeners ("ltype");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "mail" (
"id" serial NOT NULL,
"convid" bigint NOT NULL DEFAULT '0',
mail sync/migrate continued; also abstract delivery loop to make it re-usable, change refresh_all to use delivery loop.
2015-10-16 01:52:04 +00:00
"conv_guid" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"mail_flags" bigint NOT NULL DEFAULT '0',
"from_xchan" text NOT NULL DEFAULT '',
"to_xchan" text NOT NULL DEFAULT '',
"account_id" bigint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"channel_id" bigint NOT NULL DEFAULT '0',
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"mail_mimetype" varchar(64) NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"title" text NOT NULL,
"body" text NOT NULL,
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"sig" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"attach" text NOT NULL DEFAULT '',
"mid" text NOT NULL,
"parent_mid" text NOT NULL,
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"mail_deleted" smallint NOT NULL DEFAULT '0',
"mail_replied" smallint NOT NULL DEFAULT '0',
"mail_isreply" smallint NOT NULL DEFAULT '0',
"mail_seen" smallint NOT NULL DEFAULT '0',
"mail_recalled" smallint NOT NULL DEFAULT '0',
"mail_obscured" smallint NOT NULL DEFAULT '0',
database support for client side e2ee for private mail
2017-04-26 04:52:24 +00:00
"mail_raw" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"expires" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("id")
);
create index "mail_convid" on mail ("convid");
mail sync/migrate continued; also abstract delivery loop to make it re-usable, change refresh_all to use delivery loop.
2015-10-16 01:52:04 +00:00
create index "mail_conv_guid" on mail ("conv_guid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "mail_created" on mail ("created");
create index "mail_flags" on mail ("mail_flags");
create index "mail_account_id" on mail ("account_id");
create index "mail_channel_id" on mail ("channel_id");
create index "mail_from_xchan" on mail ("from_xchan");
create index "mail_to_xchan" on mail ("to_xchan");
create index "mail_mid" on mail ("mid");
create index "mail_parent_mid" on mail ("parent_mid");
create index "mail_expires" on mail ("expires");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "mail_deleted" on mail ("mail_deleted");
create index "mail_replied" on mail ("mail_replied");
create index "mail_isreply" on mail ("mail_isreply");
create index "mail_seen" on mail ("mail_seen");
create index "mail_recalled" on mail ("mail_recalled");
create index "mail_obscured" on mail ("mail_obscured");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "menu" (
"menu_id" serial NOT NULL,
"menu_channel_id" bigint NOT NULL DEFAULT '0',
"menu_name" text NOT NULL DEFAULT '',
"menu_desc" text NOT NULL DEFAULT '',
"menu_flags" bigint NOT NULL DEFAULT '0',
add menu_created, menu_edited fields to DB
2015-05-28 23:59:56 +00:00
"menu_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"menu_edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("menu_id")
);
create index "menu_channel_id" on menu ("menu_channel_id");
create index "menu_name" on menu ("menu_name");
create index "menu_flags" on menu ("menu_flags");
add menu_created, menu_edited fields to DB
2015-05-28 23:59:56 +00:00
create index "menu_created" on menu ("menu_created");
create index "menu_edited" on menu ("menu_edited");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "menu_item" (
"mitem_id" serial NOT NULL,
"mitem_link" text NOT NULL DEFAULT '',
"mitem_desc" text NOT NULL DEFAULT '',
"mitem_flags" bigint NOT NULL DEFAULT '0',
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
"mitem_channel_id" bigint NOT NULL,
"mitem_menu_id" bigint NOT NULL DEFAULT '0',
"mitem_order" bigint NOT NULL DEFAULT '0',
PRIMARY KEY ("mitem_id")
);
create index "mitem_channel_id" on menu_item ("mitem_channel_id");
create index "mitem_menu_id" on menu_item ("mitem_menu_id");
create index "mitem_flags" on menu_item ("mitem_flags");
CREATE TABLE "notify" (
"id" serial NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"hash" varchar(64) NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"xname" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"url" text NOT NULL,
"photo" text NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"created" timestamp NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"msg" text NOT NULL DEFAULT '',
"aid" bigint NOT NULL,
"uid" bigint NOT NULL,
"link" text NOT NULL,
"parent" text NOT NULL DEFAULT '',
"seen" numeric(1) NOT NULL DEFAULT '0',
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"ntype" bigint NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"verb" text NOT NULL,
"otype" varchar(16) NOT NULL,
PRIMARY KEY ("id")
);
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "notify_ntype" on notify ("ntype");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "notify_seen" on notify ("seen");
create index "notify_uid" on notify ("uid");
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "notify_created" on notify ("created");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "notify_hash" on notify ("hash");
create index "notify_parent" on notify ("parent");
create index "notify_link" on notify ("link");
create index "notify_otype" on notify ("otype");
create index "notify_aid" on notify ("aid");
CREATE TABLE "obj" (
"obj_id" serial NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"obj_page" varchar(64) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"obj_verb" text NOT NULL DEFAULT '',
typos
2016-06-27 13:28:08 +00:00
"obj_type" bigint NOT NULL DEFAULT 0,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"obj_obj" text NOT NULL DEFAULT '',
typos
2016-06-27 13:28:08 +00:00
"obj_channel" bigint NOT NULL DEFAULT 0,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"obj_term" varchar(255) NOT NULL DEFAULT '',
"obj_url" varchar(255) NOT NULL DEFAULT '',
"obj_imgurl" varchar(255) NOT NULL DEFAULT '',
separate objs/things from taxonomy
2015-09-02 00:06:58 +00:00
"obj_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"obj_edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
typos
2016-06-27 13:28:08 +00:00
"obj_quantity" bigint NOT NULL DEFAULT 0,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
PRIMARY KEY ("obj_id")
);
create index "obj_verb" on obj ("obj_verb");
create index "obj_page" on obj ("obj_page");
create index "obj_type" on obj ("obj_type");
create index "obj_channel" on obj ("obj_channel");
create index "obj_obj" on obj ("obj_obj");
separate objs/things from taxonomy
2015-09-02 00:06:58 +00:00
create index "obj_term" on obj ("obj_term");
create index "obj_url" on obj ("obj_url");
create index "obj_imgurl" on obj ("obj_imgurl");
create index "obj_created" on obj ("obj_created");
create index "obj_edited" on obj ("obj_edited");
allow objs to represent inventory
2016-05-28 06:57:47 +00:00
create index "obj_quantity" on obj ("obj_quantity");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "outq" (
"outq_hash" text NOT NULL,
"outq_account" bigint NOT NULL DEFAULT '0',
"outq_channel" bigint NOT NULL DEFAULT '0',
"outq_driver" varchar(32) NOT NULL DEFAULT '',
"outq_posturl" text NOT NULL DEFAULT '',
"outq_async" numeric(1) NOT NULL DEFAULT '0',
"outq_delivered" numeric(1) NOT NULL DEFAULT '0',
"outq_created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"outq_updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
Alter the queue so that each queue item stores the scheduled time of the next delivery. This keeps the query for queued items simple. We no longer group by posturl; as the queue update function will only keep one item per destination scheduled for shorter term processing. Others (multiple queued items for a single destination) will be scheduled for delivery far into the future and only delivered if the hub responds to the "active" or short term queue item.
2017-01-29 22:45:25 +00:00
"outq_scheduled" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"outq_notify" text NOT NULL,
"outq_msg" text NOT NULL,
add queue priority
2015-03-05 01:56:07 +00:00
"outq_priority" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("outq_hash")
);
create index "outq_account" on outq ("outq_account");
create index "outq_channel" on outq ("outq_channel");
create index "outq_hub" on outq ("outq_posturl");
create index "outq_created" on outq ("outq_created");
create index "outq_updated" on outq ("outq_updated");
Alter the queue so that each queue item stores the scheduled time of the next delivery. This keeps the query for queued items simple. We no longer group by posturl; as the queue update function will only keep one item per destination scheduled for shorter term processing. Others (multiple queued items for a single destination) will be scheduled for delivery far into the future and only delivered if the hub responds to the "active" or short term queue item.
2017-01-29 22:45:25 +00:00
create index "outq_scheduled" on outq ("outq_scheduled");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "outq_async" on outq ("outq_async");
create index "outq_delivered" on outq ("outq_delivered");
add queue priority
2015-03-05 01:56:07 +00:00
create index "outq_priority" on outq ("outq_priority");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
table structure for pseudo or proxy channels (pchan)
2017-10-20 00:24:11 +00:00
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "pconfig" (
"id" serial NOT NULL,
"uid" bigint NOT NULL DEFAULT '0',
"cat" text NOT NULL,
"k" text NOT NULL,
"v" text NOT NULL,
PRIMARY KEY ("id"),
UNIQUE ("uid","cat","k")
);
some perms fixes to stuff that was already written before I move on
2016-07-04 23:30:03 +00:00
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "photo" (
"id" serial NOT NULL,
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL,
"xchan" text NOT NULL DEFAULT '',
"resource_id" text NOT NULL,
"created" timestamp NOT NULL,
"edited" timestamp NOT NULL,
"title" text NOT NULL,
"description" text NOT NULL,
"album" text NOT NULL,
"filename" text NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"mimetype" varchar(128) NOT NULL DEFAULT 'image/jpeg',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"height" numeric(6) NOT NULL,
"width" numeric(6) NOT NULL,
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
"filesize" bigint NOT NULL DEFAULT '0',
"content" bytea NOT NULL,
"imgscale" numeric(3) NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"profile" numeric(1) NOT NULL DEFAULT '0',
updates to Postgres schema
2015-11-11 15:42:41 +00:00
"photo_usage" smallint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"is_nsfw" smallint NOT NULL DEFAULT '0',
"os_storage" smallint NOT NULL DEFAULT '0',
updates to Postgres schema
2015-11-11 15:42:41 +00:00
"os_path" text NOT NULL,
"display_path" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"photo_flags" bigint NOT NULL DEFAULT '0',
"allow_cid" text NOT NULL,
"allow_gid" text NOT NULL,
"deny_cid" text NOT NULL,
"deny_gid" text NOT NULL,
PRIMARY KEY ("id")
);
create index "photo_uid" on photo ("uid");
create index "photo_album" on photo ("album");
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "photo_imgscale" on photo ("imgscale");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "photo_profile" on photo ("profile");
create index "photo_flags" on photo ("photo_flags");
issue #511, postgres schema issues
2016-09-10 04:25:57 +00:00
create index "photo_mimetype" on photo ("mimetype");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "photo_aid" on photo ("aid");
create index "photo_xchan" on photo ("xchan");
the rest of the schema updates - WARNING: some third party plugins may fail; e.g. embedphotos and chess. $item['object'] is now $item['obj'] and $photo['type'] is $photo['mimetype'], $photo['scale'] is $photo['imgscale'] and $photo['data'] is now $photo['content']. There are a number of other changes, but these are the ones noted to cause issues with third-party plugins. The project plugins have been updated. Please note any new issues as this effort touched a lot of code in a lot of files.
2016-06-02 04:48:54 +00:00
create index "photo_filesize" on photo ("filesize");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "photo_resource_id" on photo ("resource_id");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "photo_usage" on photo ("photo_usage");
create index "photo_is_nsfw" on photo ("is_nsfw");
create index "photo_os_storage" on photo ("os_storage");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "poll" (
"poll_id" serial NOT NULL,
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
"poll_guid" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"poll_channel" bigint NOT NULL DEFAULT '0',
federated polls: add poll_author which will be necessary for Diaspora federation
2018-03-06 23:31:45 +00:00
"poll_author" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"poll_desc" text NOT NULL,
"poll_flags" bigint NOT NULL DEFAULT '0',
"poll_votes" bigint NOT NULL DEFAULT '0',
PRIMARY KEY ("poll_id")
);
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
create index "poll_guid" on poll ("poll_guid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "poll_channel" on poll ("poll_channel");
federated polls: add poll_author which will be necessary for Diaspora federation
2018-03-06 23:31:45 +00:00
create index "poll_author" on poll ("poll_author");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "poll_flags" on poll ("poll_flags");
create index "poll_votes" on poll ("poll_votes");
CREATE TABLE "poll_elm" (
"pelm_id" serial NOT NULL,
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
"pelm_guid" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"pelm_poll" bigint NOT NULL DEFAULT '0',
"pelm_desc" text NOT NULL,
"pelm_flags" bigint NOT NULL DEFAULT '0',
"pelm_result" float NOT NULL DEFAULT '0',
federated polls: GNU-Social requires us to preserve the original order of the poll answers for use in making a vote, unlike diaspora (which selects by guid) and activitypub (which selects by non-localisable text).
2018-03-07 23:39:14 +00:00
"pelm_order" numeric(6) NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("pelm_id")
);
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
create index "pelm_guid" on poll_elm ("pelm_guid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "pelm_poll" on poll_elm ("pelm_poll");
create index "pelm_result" on poll_elm ("pelm_result");
federated polls: GNU-Social requires us to preserve the original order of the poll answers for use in making a vote, unlike diaspora (which selects by guid) and activitypub (which selects by non-localisable text).
2018-03-07 23:39:14 +00:00
create index "pelm_order" on poll_elm ("pelm_order");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "profdef" (
"id" serial NOT NULL,
"field_name" text NOT NULL DEFAULT '',
"field_type" varchar(16) NOT NULL DEFAULT '',
"field_desc" text NOT NULL DEFAULT '',
"field_help" text NOT NULL DEFAULT '',
"field_inputs" text NOT NULL,
PRIMARY KEY ("id")
);
create index "profdef_field_name" on profdef ("field_name");
CREATE TABLE "profext" (
"id" serial NOT NULL,
"channel_id" bigint NOT NULL DEFAULT '0',
"hash" text NOT NULL DEFAULT '',
"k" text NOT NULL DEFAULT '',
"v" text NOT NULL,
PRIMARY KEY ("id")
);
create index "profext_channel_id" on profext ("channel_id");
create index "profext_hash" on profext ("hash");
create index "profext_k" on profext ("k");
CREATE TABLE "profile" (
"id" serial NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"profile_guid" varchar(64) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL,
"profile_name" text NOT NULL,
"is_default" numeric(1) NOT NULL DEFAULT '0',
"hide_friends" numeric(1) NOT NULL DEFAULT '0',
more db column renames
2016-06-01 04:45:33 +00:00
"fullname" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"pdesc" text NOT NULL DEFAULT '',
"chandesc" text NOT NULL DEFAULT '',
"dob" varchar(32) NOT NULL DEFAULT '',
"dob_tz" text NOT NULL DEFAULT 'UTC',
"address" text NOT NULL DEFAULT '',
"locality" text NOT NULL DEFAULT '',
"region" text NOT NULL DEFAULT '',
"postal_code" varchar(32) NOT NULL DEFAULT '',
"country_name" text NOT NULL DEFAULT '',
"hometown" text NOT NULL DEFAULT '',
"gender" varchar(32) NOT NULL DEFAULT '',
"marital" text NOT NULL DEFAULT '',
more db column renames
2016-06-01 04:45:33 +00:00
"partner" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"howlong" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"sexual" text NOT NULL DEFAULT '',
"politic" text NOT NULL DEFAULT '',
"religion" text NOT NULL DEFAULT '',
"keywords" text NOT NULL DEFAULT '',
"likes" text NOT NULL DEFAULT '',
"dislikes" text NOT NULL DEFAULT '',
"about" text NOT NULL DEFAULT '',
"summary" text NOT NULL DEFAULT '',
"music" text NOT NULL DEFAULT '',
"book" text NOT NULL DEFAULT '',
"tv" text NOT NULL DEFAULT '',
"film" text NOT NULL DEFAULT '',
"interest" text NOT NULL DEFAULT '',
"romance" text NOT NULL DEFAULT '',
more db column renames
2016-06-01 04:45:33 +00:00
"employment" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"education" text NOT NULL DEFAULT '',
"contact" text NOT NULL DEFAULT '',
"channels" text NOT NULL DEFAULT '',
"homepage" text NOT NULL DEFAULT '',
"photo" text NOT NULL,
"thumb" text NOT NULL,
"publish" numeric(1) NOT NULL DEFAULT '0',
more vcard work
2016-12-21 23:03:33 +00:00
"profile_vcard" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id"),
UNIQUE ("profile_guid","uid")
);
create index "profile_uid" on profile ("uid");
create index "profile_locality" on profile ("locality");
create index "profile_hometown" on profile ("hometown");
create index "profile_gender" on profile ("gender");
create index "profile_marital" on profile ("marital");
create index "profile_sexual" on profile ("sexual");
create index "profile_publish" on profile ("publish");
create index "profile_aid" on profile ("aid");
create index "profile_is_default" on profile ("is_default");
create index "profile_hide_friends" on profile ("hide_friends");
create index "profile_postal_code" on profile ("postal_code");
create index "profile_country_name" on profile ("country_name");
create index "profile_guid" on profile ("profile_guid");
CREATE TABLE "profile_check" (
"id" serial NOT NULL,
"uid" bigint NOT NULL,
"cid" bigint NOT NULL DEFAULT '0',
"dfrn_id" text NOT NULL,
"sec" text NOT NULL,
"expire" bigint NOT NULL,
PRIMARY KEY ("id")
);
create index "pc_uid" on profile_check ("uid");
create index "pc_cid" on profile_check ("cid");
create index "pc_dfrn_id" on profile_check ("dfrn_id");
create index "pc_sec" on profile_check ("sec");
create index "pc_expire" on profile_check ("expire");
CREATE TABLE "register" (
"id" serial NOT NULL,
"hash" text NOT NULL,
"created" timestamp NOT NULL,
"uid" bigint NOT NULL,
"password" text NOT NULL,
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"lang" varchar(16) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("id")
);
create index "reg_hash" on register ("hash");
create index "reg_created" on register ("created");
create index "reg_uid" on register ("uid");
CREATE TABLE "session" (
"id" serial,
"sid" text NOT NULL,
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"sess_data" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"expire" numeric(20) NOT NULL,
PRIMARY KEY ("id")
);
create index "session_sid" on session ("sid");
create index "session_expire" on session ("expire");
CREATE TABLE "shares" (
"share_id" serial NOT NULL,
"share_type" bigint NOT NULL DEFAULT '0',
"share_target" bigint NOT NULL DEFAULT '0',
"share_xchan" text NOT NULL DEFAULT '',
PRIMARY KEY ("share_id")
);
create index "share_type" on shares ("share_type");
create index "share_target" on shares ("share_target");
create index "share_xchan" on shares ("share_xchan");
CREATE TABLE "sign" (
"id" serial NOT NULL,
"iid" bigint NOT NULL DEFAULT '0',
"retract_iid" bigint NOT NULL DEFAULT '0',
"signed_text" text NOT NULL,
"signature" text NOT NULL,
"signer" text NOT NULL,
PRIMARY KEY ("id")
);
create index "sign_iid" on "sign" ("iid");
create index "sign_retract_iid" on "sign" ("retract_iid");
CREATE TABLE "site" (
"site_url" text NOT NULL,
"site_access" bigint NOT NULL DEFAULT '0',
"site_flags" bigint NOT NULL DEFAULT '0',
"site_update" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"site_pull" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"site_sync" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"site_directory" text NOT NULL DEFAULT '',
"site_register" bigint NOT NULL DEFAULT '0',
"site_sellpage" text NOT NULL DEFAULT '',
"site_location" text NOT NULL DEFAULT '',
"site_realm" text NOT NULL DEFAULT '',
require access token to view, query, or join directories in private realms, if the realm is so configured.
2015-02-25 00:36:27 +00:00
"site_valid" smallint NOT NULL DEFAULT '0',
add site_dead flag to prevent delivery to dead sites. Allow sys channel webpages to be viewed even if site is configured "block public".
2015-06-08 23:55:17 +00:00
"site_dead" smallint NOT NULL DEFAULT '0',
schema support for additional site types
2015-09-28 11:15:21 +00:00
"site_type" smallint NOT NULL DEFAULT '0',
add the project name to site info packet
2015-10-19 23:07:14 +00:00
"site_project" text NOT NULL DEFAULT '',
provide version info in /pubsites
2016-09-08 00:36:45 +00:00
"site_version" text NOT NULL DEFAULT '',
db update required for Zot2
2016-11-22 22:15:33 +00:00
"site_crypto" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("site_url")
);
create index "site_flags" on site ("site_flags");
create index "site_update" on site ("site_update");
create index "site_directory" on site ("site_directory");
create index "site_register" on site ("site_register");
create index "site_access" on site ("site_access");
create index "site_sellpage" on site ("site_sellpage");
create index "site_realm" on site ("site_realm");
require access token to view, query, or join directories in private realms, if the realm is so configured.
2015-02-25 00:36:27 +00:00
create index "site_valid" on site ("site_valid");
add site_dead flag to prevent delivery to dead sites. Allow sys channel webpages to be viewed even if site is configured "block public".
2015-06-08 23:55:17 +00:00
create index "site_dead" on site ("site_dead");
schema support for additional site types
2015-09-28 11:15:21 +00:00
create index "site_type" on site ("site_type");
add the project name to site info packet
2015-10-19 23:07:14 +00:00
create index "site_project" on site ("site_project");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "source" (
"src_id" serial NOT NULL,
"src_channel_id" bigint NOT NULL DEFAULT '0',
"src_channel_xchan" text NOT NULL DEFAULT '',
"src_xchan" text NOT NULL DEFAULT '',
db schema change to add tags to content sources
2016-05-02 04:00:02 +00:00
"src_patt" text NOT NULL DEFAULT '',
"src_tag" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("src_id")
);
create index "src_channel_id" on "source" ("src_channel_id");
create index "src_channel_xchan" on "source" ("src_channel_xchan");
create index "src_xchan" on "source" ("src_xchan");
CREATE TABLE "sys_perms" (
"id" serial NOT NULL,
"cat" text NOT NULL,
"k" text NOT NULL,
"v" text NOT NULL,
"public_perm" numeric(1) NOT NULL,
PRIMARY KEY ("id")
);
CREATE TABLE "term" (
"tid" serial NOT NULL,
"aid" bigint NOT NULL DEFAULT '0',
"uid" bigint NOT NULL DEFAULT '0',
"oid" bigint NOT NULL,
"otype" numeric(3) NOT NULL,
more db column renames
2016-06-01 04:45:33 +00:00
"ttype" numeric(3) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"term" text NOT NULL,
"url" text NOT NULL,
"imgurl" text NOT NULL DEFAULT '',
"term_hash" text NOT NULL DEFAULT '',
"parent_hash" text NOT NULL DEFAULT '',
PRIMARY KEY ("tid")
);
create index "term_oid" on term ("oid");
create index "term_otype" on term ("otype");
more db column renames
2016-06-01 04:45:33 +00:00
create index "term_ttype" on term ("ttype");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "term_term" on term ("term");
create index "term_uid" on term ("uid");
create index "term_aid" on term ("aid");
create index "term_imgurl" on term ("imgurl");
create index "term_hash" on term ("term_hash");
create index "term_parent_hash" on term ("parent_hash");
CREATE TABLE "tokens" (
"id" varchar(40) NOT NULL,
"secret" text NOT NULL,
"client_id" varchar(20) NOT NULL,
"expires" numeric(20) NOT NULL,
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"auth_scope" varchar(512) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"uid" bigint NOT NULL,
PRIMARY KEY ("id")
);
create index "tokens_client_id" on tokens ("client_id");
create index "tokens_expires" on tokens ("expires");
create index "tokens_uid" on tokens ("uid");
CREATE TABLE "updates" (
"ud_id" serial NOT NULL,
danger - do not use 'char(n)' with postgres it creates a thing called a bpchar which is different than a normal text field in subtle ways.
2017-05-16 05:58:13 +00:00
"ud_hash" varchar(128) NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"ud_guid" text NOT NULL DEFAULT '',
"ud_date" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"ud_last" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"ud_flags" bigint NOT NULL DEFAULT '0',
"ud_addr" text NOT NULL DEFAULT '',
PRIMARY KEY ("ud_id")
);
create index "ud_date" on updates ("ud_date");
create index "ud_guid" on updates ("ud_guid");
create index "ud_hash" on updates ("ud_hash");
create index "ud_flags" on updates ("ud_flags");
create index "ud_addr" on updates ("ud_addr");
create index "ud_last" on updates ("ud_last");
CREATE TABLE "verify" (
"id" serial NOT NULL,
"channel" bigint NOT NULL DEFAULT '0',
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
"vtype" varchar(32) NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"token" text NOT NULL DEFAULT '',
"meta" text NOT NULL DEFAULT '',
"created" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("id")
);
create index "verify_channel" on verify ("channel");
more removal of reserved words from DB schemas
2016-06-01 00:50:47 +00:00
create index "verify_vtype" on verify ("vtype");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "verify_token" on verify ("token");
create index "verify_meta" on verify ("meta");
create index "verify_created" on verify ("created");
CREATE TABLE "vote" (
"vote_id" serial NOT NULL,
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
"vote_guid" text NOT NULL,
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"vote_poll" bigint NOT NULL DEFAULT '0',
"vote_element" bigint NOT NULL DEFAULT '0',
"vote_result" text NOT NULL,
"vote_xchan" text NOT NULL DEFAULT '',
PRIMARY KEY ("vote_id"),
UNIQUE ("vote_poll","vote_element","vote_xchan")
);
provide DB compatibility for poll and voting implementations across several platforms
2018-02-23 04:52:39 +00:00
create index "vote_guid" on vote ("vote_guid");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
create index "vote_poll" on vote ("vote_poll");
create index "vote_element" on vote ("vote_element");
CREATE TABLE "xchan" (
"xchan_hash" text NOT NULL,
"xchan_guid" text NOT NULL DEFAULT '',
"xchan_guid_sig" text NOT NULL DEFAULT '',
"xchan_pubkey" text NOT NULL DEFAULT '',
"xchan_photo_mimetype" text NOT NULL DEFAULT 'image/jpeg',
"xchan_photo_l" text NOT NULL DEFAULT '',
"xchan_photo_m" text NOT NULL DEFAULT '',
"xchan_photo_s" text NOT NULL DEFAULT '',
"xchan_addr" text NOT NULL DEFAULT '',
"xchan_url" text NOT NULL DEFAULT '',
"xchan_connurl" text NOT NULL DEFAULT '',
"xchan_follow" text NOT NULL DEFAULT '',
"xchan_connpage" text NOT NULL DEFAULT '',
"xchan_name" text NOT NULL DEFAULT '',
"xchan_network" text NOT NULL DEFAULT '',
"xchan_instance_url" text NOT NULL DEFAULT '',
"xchan_flags" bigint NOT NULL DEFAULT '0',
"xchan_photo_date" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
"xchan_name_date" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"xchan_hidden" smallint NOT NULL DEFAULT '0',
"xchan_orphan" smallint NOT NULL DEFAULT '0',
"xchan_censored" smallint NOT NULL DEFAULT '0',
"xchan_selfcensored" smallint NOT NULL DEFAULT '0',
"xchan_system" smallint NOT NULL DEFAULT '0',
begin change xchan_pubforum to xchan_type
2019-03-05 23:35:17 +00:00
"xchan_type" smallint NOT NULL DEFAULT '0',
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
"xchan_deleted" smallint NOT NULL DEFAULT '0',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("xchan_hash")
);
create index "xchan_guid" on xchan ("xchan_guid");
create index "xchan_addr" on xchan ("xchan_addr");
create index "xchan_name" on xchan ("xchan_name");
create index "xchan_network" on xchan ("xchan_network");
create index "xchan_url" on xchan ("xchan_url");
create index "xchan_flags" on xchan ("xchan_flags");
create index "xchan_connurl" on xchan ("xchan_connurl");
create index "xchan_instance_url" on xchan ("xchan_instance_url");
create index "xchan_follow" on xchan ("xchan_follow");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "xchan_hidden" on xchan ("xchan_hidden");
create index "xchan_orphan" on xchan ("xchan_orphan");
create index "xchan_censored" on xchan ("xchan_censored");
create index "xchan_selfcensored" on xchan ("xchan_selfcensored");
create index "xchan_system" on xchan ("xchan_system");
begin change xchan_pubforum to xchan_type
2019-03-05 23:35:17 +00:00
create index "xchan_type" on xchan ("xchan_type");
fix the postgres schema with all the changes to date
2015-06-30 02:58:44 +00:00
create index "xchan_deleted" on xchan ("xchan_deleted");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "xchat" (
"xchat_id" serial NOT NULL,
"xchat_url" text NOT NULL DEFAULT '',
"xchat_desc" text NOT NULL DEFAULT '',
"xchat_xchan" text NOT NULL DEFAULT '',
"xchat_edited" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
PRIMARY KEY ("xchat_id")
);
create index "xchat_url" on xchat ("xchat_url");
create index "xchat_desc" on xchat ("xchat_desc");
create index "xchat_xchan" on xchat ("xchat_xchan");
create index "xchat_edited" on xchat ("xchat_edited");
CREATE TABLE "xconfig" (
"id" serial NOT NULL,
"xchan" text NOT NULL,
"cat" text NOT NULL,
"k" text NOT NULL,
"v" text NOT NULL,
PRIMARY KEY ("id")
);
create index "xconfig_xchan" on xconfig ("xchan");
create index "xconfig_cat" on xconfig ("cat");
create index "xconfig_k" on xconfig ("k");
CREATE TABLE "xign" (
"id" serial NOT NULL,
"uid" bigint NOT NULL DEFAULT '0',
"xchan" text NOT NULL DEFAULT '',
PRIMARY KEY ("id")
);
create index "xign_uid" on xign ("uid");
create index "xign_xchan" on xign ("xchan");
CREATE TABLE "xlink" (
"xlink_id" serial NOT NULL,
"xlink_xchan" text NOT NULL DEFAULT '',
"xlink_link" text NOT NULL DEFAULT '',
"xlink_rating" bigint NOT NULL DEFAULT '0',
add db update 1131/1132 to database schemas
2015-01-23 22:20:30 +00:00
"xlink_rating_text" TEXT NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
"xlink_updated" timestamp NOT NULL DEFAULT '0001-01-01 00:00:00',
provide storage for directory based reputation in the xlink table by setting xlink_static = 1, so that xlink_static = 0 is traditional poco linkages
2015-01-27 03:13:06 +00:00
"xlink_static" numeric(1) NOT NULL DEFAULT '0',
two tracks proceeding in parallel - first adding the ability to store a digital signature with ratings so that directories can pass them around, second provide some directory registration functions so we can obtain a list of directories from somewhere (to pass ratings around between them). This gives the primary directory role some value as that is where you register your directory.
2015-02-01 23:10:52 +00:00
"xlink_sig" text NOT NULL DEFAULT '',
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
PRIMARY KEY ("xlink_id")
);
create index "xlink_xchan" on xlink ("xlink_xchan");
create index "xlink_link" on xlink ("xlink_link");
create index "xlink_updated" on xlink ("xlink_updated");
create index "xlink_rating" on xlink ("xlink_rating");
provide storage for directory based reputation in the xlink table by setting xlink_static = 1, so that xlink_static = 0 is traditional poco linkages
2015-01-27 03:13:06 +00:00
create index "xlink_static" on xlink ("xlink_static");
oauth permissions table
2015-01-20 04:06:12 +00:00
CREATE TABLE "xperm" (
"xp_id" serial NOT NULL,
"xp_client" varchar( 20 ) NOT NULL DEFAULT '',
"xp_channel" bigint NOT NULL DEFAULT '0',
"xp_perm" varchar( 64 ) NOT NULL DEFAULT '',
fix issues in schema_postgres.sql - change PRIMARY_KEY to PRIMARY KEY - replace smallint(1) with smallint
2015-02-26 22:50:46 +00:00
PRIMARY KEY ("xp_id")
oauth permissions table
2015-01-20 04:06:12 +00:00
);
create index "xp_client" on xperm ("xp_client");
create index "xp_channel" on xperm ("xp_channel");
create index "xp_perm" on xperm ("xp_perm");
PostgreSQL support initial commit There were 11 main types of changes: - UPDATE's and DELETE's sometimes had LIMIT 1 at the end of them. This is not only non-compliant but it would certainly not do what whoever wrote it thought it would. It is likely this mistake was just copied from Friendica. All of these instances, the LIMIT 1 was simply removed. - Bitwise operations (and even some non-zero int checks) erroneously rely on MySQL implicit integer-boolean conversion in the WHERE clauses. This is non-compliant (and bad programming practice to boot). Proper explicit boolean conversions were added. New queries should use proper conventions. - MySQL has a different operator for bitwise XOR than postgres. Rather than add yet another dba_ func, I converted them to "& ~" ("AND NOT") when turning off, and "|" ("OR") when turning on. There were no true toggles (XOR). New queries should refrain from using XOR when not necessary. - There are several fields which the schema has marked as NOT NULL, but the inserts don't specify them. The reason this works is because mysql totally ignores the constraint and adds an empty text default automatically. Again, non-compliant, obviously. In these cases a default of empty text was added. - Several statements rely on a non-standard MySQL feature (http://dev.mysql.com/doc/refman/5.5/en/group-by-handling.html). These queries can all be rewritten to be standards compliant. Interestingly enough, the newly rewritten standards compliant queries run a zillion times faster, even on MySQL. - A couple of function/operator name translations were needed (RAND/RANDOM, GROUP_CONCAT/STRING_AGG, UTC_NOW, REGEXP/~, ^/#) -- assist functions added in the dba_ - INTERVALs: postgres requires quotes around the value, mysql requires that there are not quotes around the value -- assist functions added in the dba_ - NULL_DATE's -- Postgres does not allow the invalid date '0000-00-00 00:00:00' (there is no such thing as year 0 or month 0 or day 0). We use '0001-01-01 00:00:00' for postgres. Conversions are handled in Zot/item packets automagically by quoting all dates with dbescdate(). - char(##) specifications in the schema creates fields with blank spaces that aren't trimmed in the code. MySQL apparently treats char(##) as varchar(##), again, non-compliant. Since postgres works better with text fields anyway, this ball of bugs was simply side-stepped by using 'text' datatype for all text fields in the postgres schema. varchar was used in a couple of places where it actually seemed appropriate (size constraint), but without rigorously vetting that all of the PHP code actually validates data, new bugs might come out from under the rug. - postgres doesn't store nul bytes and a few other non-printables in text fields, even when quoted. bytea fields were used when storing binary data (photo.data, attach.data). A new dbescbin() function was added to handle this transparently. - postgres does not support LIMIT #,# syntax. All databases support LIMIT # OFFSET # syntax. Statements were updated to be standard. These changes require corresponding changes in the coding standards. Please review those before adding any code going forward. Still on my TODO list: - remove quotes from non-reserved identifiers and make reserved identifiers use dba func for quoting - Rewrite search queries for better results (both MySQL and Postgres)
2014-11-13 20:21:58 +00:00
CREATE TABLE "xprof" (
"xprof_hash" text NOT NULL,
"xprof_age" numeric(3) NOT NULL DEFAULT '0',
"xprof_desc" text NOT NULL DEFAULT '',
"xprof_dob" varchar(12) NOT NULL DEFAULT '',
"xprof_gender" text NOT NULL DEFAULT '',
"xprof_marital" text NOT NULL DEFAULT '',
"xprof_sexual" text NOT NULL DEFAULT '',
"xprof_locale" text NOT NULL DEFAULT '',
"xprof_region" text NOT NULL DEFAULT '',
"xprof_postcode" varchar(32) NOT NULL DEFAULT '',
"xprof_country" text NOT NULL DEFAULT '',
"xprof_keywords" text NOT NULL,
"xprof_about" text NOT NULL,
"xprof_homepage" text NOT NULL DEFAULT '',
"xprof_hometown" text NOT NULL DEFAULT '',
PRIMARY KEY ("xprof_hash")
);
create index "xprof_desc" on xprof ("xprof_desc");
create index "xprof_dob" on xprof ("xprof_dob");
create index "xprof_gender" on xprof ("xprof_gender");
create index "xprof_marital" on xprof ("xprof_marital");
create index "xprof_sexual" on xprof ("xprof_sexual");
create index "xprof_locale" on xprof ("xprof_locale");
create index "xprof_region" on xprof ("xprof_region");
create index "xprof_postcode" on xprof ("xprof_postcode");
create index "xprof_country" on xprof ("xprof_country");
create index "xprof_age" on xprof ("xprof_age");
create index "xprof_hometown" on xprof ("xprof_hometown");
CREATE TABLE "xtag" (
"xtag_id" serial NOT NULL,
"xtag_hash" text NOT NULL,
"xtag_term" text NOT NULL DEFAULT '',
"xtag_flags" bigint NOT NULL DEFAULT '0',
PRIMARY KEY ("xtag_id")
);
create index "xtag_term" on xtag ("xtag_term");
create index "xtag_hash" on xtag ("xtag_hash");
create index "xtag_flags" on xtag ("xtag_flags");
add the cdav pgsql schemas
2017-07-04 08:28:16 +00:00
CREATE TABLE addressbooks (
id SERIAL NOT NULL,
principaluri VARCHAR(255),
displayname VARCHAR(255),
uri VARCHAR(200),
description TEXT,
synctoken INTEGER NOT NULL DEFAULT 1
);
ALTER TABLE ONLY addressbooks
ADD CONSTRAINT addressbooks_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX addressbooks_ukey
ON addressbooks USING btree (principaluri, uri);
CREATE TABLE cards (
id SERIAL NOT NULL,
addressbookid INTEGER NOT NULL,
carddata BYTEA,
uri VARCHAR(200),
lastmodified INTEGER,
etag VARCHAR(32),
size INTEGER NOT NULL
);
ALTER TABLE ONLY cards
ADD CONSTRAINT cards_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX cards_ukey
ON cards USING btree (addressbookid, uri);
CREATE TABLE addressbookchanges (
id SERIAL NOT NULL,
uri VARCHAR(200) NOT NULL,
synctoken INTEGER NOT NULL,
addressbookid INTEGER NOT NULL,
operation SMALLINT NOT NULL
);
ALTER TABLE ONLY addressbookchanges
ADD CONSTRAINT addressbookchanges_pkey PRIMARY KEY (id);
CREATE INDEX addressbookchanges_addressbookid_synctoken_ix
ON addressbookchanges USING btree (addressbookid, synctoken);
CREATE TABLE calendarobjects (
id SERIAL NOT NULL,
calendardata BYTEA,
uri VARCHAR(200),
calendarid INTEGER NOT NULL,
lastmodified INTEGER,
etag VARCHAR(32),
size INTEGER NOT NULL,
componenttype VARCHAR(8),
firstoccurence INTEGER,
lastoccurence INTEGER,
uid VARCHAR(200)
);
ALTER TABLE ONLY calendarobjects
ADD CONSTRAINT calendarobjects_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX calendarobjects_ukey
ON calendarobjects USING btree (calendarid, uri);
CREATE TABLE calendars (
id SERIAL NOT NULL,
synctoken INTEGER NOT NULL DEFAULT 1,
components VARCHAR(21)
);
ALTER TABLE ONLY calendars
ADD CONSTRAINT calendars_pkey PRIMARY KEY (id);
CREATE TABLE calendarinstances (
id SERIAL NOT NULL,
calendarid INTEGER NOT NULL,
principaluri VARCHAR(100),
access SMALLINT NOT NULL DEFAULT '1', -- '1 = owner, 2 = read, 3 = readwrite'
displayname VARCHAR(100),
uri VARCHAR(200),
description TEXT,
calendarorder INTEGER NOT NULL DEFAULT 0,
calendarcolor VARCHAR(10),
timezone TEXT,
transparent SMALLINT NOT NULL DEFAULT '0',
share_href VARCHAR(100),
share_displayname VARCHAR(100),
share_invitestatus SMALLINT NOT NULL DEFAULT '2' -- '1 = noresponse, 2 = accepted, 3 = declined, 4 = invalid'
);
ALTER TABLE ONLY calendarinstances
ADD CONSTRAINT calendarinstances_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX calendarinstances_principaluri_uri
ON calendarinstances USING btree (principaluri, uri);
CREATE UNIQUE INDEX calendarinstances_principaluri_calendarid
ON calendarinstances USING btree (principaluri, calendarid);
CREATE UNIQUE INDEX calendarinstances_principaluri_share_href
ON calendarinstances USING btree (principaluri, share_href);
CREATE TABLE calendarsubscriptions (
id SERIAL NOT NULL,
uri VARCHAR(200) NOT NULL,
principaluri VARCHAR(100) NOT NULL,
source TEXT,
displayname VARCHAR(100),
refreshrate VARCHAR(10),
calendarorder INTEGER NOT NULL DEFAULT 0,
calendarcolor VARCHAR(10),
striptodos SMALLINT NULL,
stripalarms SMALLINT NULL,
stripattachments SMALLINT NULL,
lastmodified INTEGER
);
ALTER TABLE ONLY calendarsubscriptions
ADD CONSTRAINT calendarsubscriptions_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX calendarsubscriptions_ukey
ON calendarsubscriptions USING btree (principaluri, uri);
CREATE TABLE calendarchanges (
id SERIAL NOT NULL,
uri VARCHAR(200) NOT NULL,
synctoken INTEGER NOT NULL,
calendarid INTEGER NOT NULL,
operation SMALLINT NOT NULL DEFAULT 0
);
ALTER TABLE ONLY calendarchanges
ADD CONSTRAINT calendarchanges_pkey PRIMARY KEY (id);
CREATE INDEX calendarchanges_calendarid_synctoken_ix
ON calendarchanges USING btree (calendarid, synctoken);
CREATE TABLE schedulingobjects (
id SERIAL NOT NULL,
principaluri VARCHAR(255),
calendardata BYTEA,
uri VARCHAR(200),
lastmodified INTEGER,
etag VARCHAR(32),
size INTEGER NOT NULL
);
CREATE TABLE locks (
id SERIAL NOT NULL,
owner VARCHAR(100),
timeout INTEGER,
created INTEGER,
token VARCHAR(100),
scope SMALLINT,
depth SMALLINT,
uri TEXT
);
ALTER TABLE ONLY locks
ADD CONSTRAINT locks_pkey PRIMARY KEY (id);
CREATE INDEX locks_token_ix
ON locks USING btree (token);
CREATE INDEX locks_uri_ix
ON locks USING btree (uri);
CREATE TABLE principals (
id SERIAL NOT NULL,
uri VARCHAR(200) NOT NULL,
email VARCHAR(80),
displayname VARCHAR(80)
);
ALTER TABLE ONLY principals
ADD CONSTRAINT principals_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX principals_ukey
ON principals USING btree (uri);
CREATE TABLE groupmembers (
id SERIAL NOT NULL,
principal_id INTEGER NOT NULL,
member_id INTEGER NOT NULL
);
ALTER TABLE ONLY groupmembers
ADD CONSTRAINT groupmembers_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX groupmembers_ukey
ON groupmembers USING btree (principal_id, member_id);
CREATE TABLE propertystorage (
id SERIAL NOT NULL,
path VARCHAR(1024) NOT NULL,
name VARCHAR(100) NOT NULL,
valuetype INT,
value BYTEA
);
ALTER TABLE ONLY propertystorage
ADD CONSTRAINT propertystorage_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX propertystorage_ukey
ON propertystorage (path, name);
CREATE TABLE users (
id SERIAL NOT NULL,
username VARCHAR(50),
digesta1 VARCHAR(32)
);
ALTER TABLE ONLY users
ADD CONSTRAINT users_pkey PRIMARY KEY (id);
CREATE UNIQUE INDEX users_ukey
ON users USING btree (username);
OAuth2 integration
2018-04-04 04:08:40 +00:00
CREATE TABLE oauth_clients (
client_id VARCHAR(80) NOT NULL,
client_secret VARCHAR(80),
redirect_uri VARCHAR(2000),
grant_types VARCHAR(80),
scope VARCHAR(4000),
update oauth related tables to use bigint/int(10) for user_id column. this is to be more consistent with the rest of the tables and fixes issue #1180
2018-05-20 20:42:47 +00:00
user_id bigint NOT NULL DEFAULT '0',
OAuth2 integration
2018-04-04 04:08:40 +00:00
PRIMARY KEY (client_id)
);
CREATE TABLE oauth_access_tokens (
access_token VARCHAR(40) NOT NULL,
client_id VARCHAR(80) NOT NULL,
update oauth related tables to use bigint/int(10) for user_id column. this is to be more consistent with the rest of the tables and fixes issue #1180
2018-05-20 20:42:47 +00:00
user_id bigint NOT NULL DEFAULT '0',
OAuth2 integration
2018-04-04 04:08:40 +00:00
expires TIMESTAMP NOT NULL,
scope VARCHAR(4000),
PRIMARY KEY (access_token)
);
CREATE TABLE oauth_authorization_codes (
authorization_code VARCHAR(40) NOT NULL,
client_id VARCHAR(80) NOT NULL,
update oauth related tables to use bigint/int(10) for user_id column. this is to be more consistent with the rest of the tables and fixes issue #1180
2018-05-20 20:42:47 +00:00
user_id bigint NOT NULL DEFAULT '0',
OAuth2 integration
2018-04-04 04:08:40 +00:00
redirect_uri VARCHAR(2000),
expires TIMESTAMP NOT NULL,
scope VARCHAR(4000),
id_token VARCHAR(1000),
PRIMARY KEY (authorization_code)
);
CREATE TABLE oauth_refresh_tokens (
refresh_token VARCHAR(40) NOT NULL,
client_id VARCHAR(80) NOT NULL,
update oauth related tables to use bigint/int(10) for user_id column. this is to be more consistent with the rest of the tables and fixes issue #1180
2018-05-20 20:42:47 +00:00
user_id bigint NOT NULL DEFAULT '0',
OAuth2 integration
2018-04-04 04:08:40 +00:00
expires TIMESTAMP NOT NULL,
scope VARCHAR(4000),
PRIMARY KEY (refresh_token)
);
CREATE TABLE oauth_scopes (
scope VARCHAR(191) NOT NULL,
is_default SMALLINT,
PRIMARY KEY (scope)
);
CREATE TABLE oauth_jwt (
client_id VARCHAR(80) NOT NULL,
subject VARCHAR(80),
public_key VARCHAR(2000) NOT NULL
);
Reference in a new issue Copy permalink