Make IDatabase::lastDoneWrites() reflect creation and changes to
the cloned temporary unit test tables but not other temporary tables.
This effects the LB method hasOrMadeRecentMasterChanges(). Other tables
are assumpted to really just be there for temporary calculations rather
acting as test-only ephemeral versions of permanent tables. Treating
writes to the "fake permanent" temp tables more like real permanent
tables means that the tests better align with production.
At the moment, temporary tables still have to use DB_MASTER, given
the assertIsWritableMaster() check in query(). This restriction
can be lifted at some point, when RDBMs compatibility is robust.
Bug: T218388
Change-Id: I4c0d629da254ac2aaf31aae35bd2efc7bc064ac6
Also:
* Fixed LoadBalancer::getAnyOpenConnection for both
DB_MASTER and DB_REPLICA, which are not real indexes.
* Lock down DBConnRef::close since it can only cause trouble.
* Relax DBConnRef restrictions on tablePrefix()/dbSchema()
for the harmless "getter" mode case.
* Remove redundant DatabasePostgres::getServer definition.
Change-Id: Ia855d901cc3c28147e52284fdabb1645805d4466
PostgreSQL doesn't support anything like this. For now, avoid generating
invalid SQL by just ignoring the option. If we come up with a use case
someday, that can guide implementation of a workalike.
Also, remove a pointless "IGNORE" from populateExternallinksIndex60.php.
el_index_60 isn't uniquely indexed, so it has no effect anyway.
Bug: T215169
Change-Id: I1409c80b39834d1977c82c489226255a8cc93fd0
PostgreSQL v12 will remove the long-deprecated column
pg_attrdef.adsrc. The supported way to introspect into column
default values is pg_get_expr(adbin, adrelid), which works
back through all versions of PostgreSQL supported by wikimedia.
Changing to the supported method will allow the upcoming v12 of the
database to be used while maintaining compatibility with older
versions, without needing to write version-specific code.
This patch has been tested with maintenance/update.php and
with phpunit in PostgreSQL versions 9.2, 11, and 12dev. It does
not harm the first two, and fixes errors that would otherwise
arise in the dev version. All unit tests which pass under version
11 now pass under 12dev as well.
Change-Id: I874d347fd286b26773113d4f0c6c30d9a4055ad3
Also improved the atomicity and affected row count logic for
insert/replace with sqlite.
Also remove unused "fileHandle" code from insert().
Change-Id: If7b9148fd44f3a958899885753c7c86ba66bf193
LoadBalancer uses Database::getDomainId() for deciding which keys to use
in the foreign connection handle arrays. This method should reflect any
changes made to the DB selection.
If the query fails, then do not change domain field. This is the sort of
approach that LoadBalancer is expects in openForeignConnection(). Also,
throw an exception when selectDB() fails.
The db/schema/prefix fields of Database no longer exist in favor of just
using the newer currentDomain field.
Also:
* Add IDatabase::selectDomain() method and made selectDB() wrap it.
* Extract the DB name from sqlite files if not explicitly provided.
* Fix inconsistent open() return values from Database subclasses.
* Make a relationSchemaQualifier() method to handle the concern of
omitting schema names in queries. The means that getDomainId() can
still return the right value, rather than confusingly omitt the schema.
* Make RevisionStore::checkDatabaseWikiId() account for the domain schema.
Unlike d2a4d614fc, this does not incorrectly assume the storage is
always for the current wiki domain. Also, LBFactorySingle sets the local
domain so it is defined even in install.php.
* Make RevisionStoreDbTestBase actually set the LoadBalancer local domain.
* Make RevisionTest::testLoadFromTitle() account for the domain schema.
Bug: T193565
Change-Id: I6e51cd54c6da78830b38906b8c46789c79498ab5
This is not called externally and there is little reason for that to
change. The current caller pattern is to use factory(), possibly with
initConnection() afterwards, or to use a LoadBalancer to begin with.
Change-Id: Ib1fdd5c960f1ed877fcd17bcb99b999d5d894716
Find: /isset\(\s*([^()]+?)\s*\)\s*\?\s*\1\s*:\s*/
Replace with: '\1 ?? '
(Everywhere except includes/PHPVersionCheck.php)
(Then, manually fix some line length and indentation issues)
Then manually reviewed the replacements for cases where confusing
operator precedence would result in incorrect results
(fixing those in I478db046a1cc162c6767003ce45c9b56270f3372).
Change-Id: I33b421c8cb11cdd4ce896488c9ff5313f03a38cf
PostgreSQL allows setting an "owner" column for a sequence, so if that
column is dropped then the sequence will be dropped too. We should
certainly take advantage of that when creating duplicate tables for unit
testing (particularly when $temporary is false), and we may as well do
it for our permanent tables too.
Change-Id: I4822ac33298e3f3ef59f4372a24aa0866a6e66ae
This improves the repeatability of the unit tests by making the ID
values generated depend less on what previous tests might have done.
It also prevents tests from using up sequence numbers for the live DB's
tables.
Change-Id: Iaa8ae1e5cef4b9099bd1b4b8fc806f5af372a7ff
PostgreSQL puts temporary tables and such in a hidden, per-connection
"schema" that's checked for unqualified table accesses before the normal
search_path. We should check that in all the schema-checking functions.
Change-Id: I1194ac31f31133b177f624138afee19d00e454b9
MediaWiki doesn't support PostgreSQL < 9.2, so drop the support for
older versions.
At the same time, since we're messing with the DatabasePostgres::insert()
code anyway, let's start using ON CONFLICT DO NOTHING for PG >= 9.5.
And since we're doing that, let's do the same for
DatabasePostgres::nativeInsertSelect().
Change-Id: I7bf13c3272917ebafeaff11eb116714a099afdf3
Handle all errors in query() that might have caused rollback by
putting the Database handle into an error state that can only be
resolved by cancelAtomic() or rollback(). Other queries will be
rejected until then.
This results in more immediate exceptions in some cases where
atomic section mismatch errors would have been thrown, such as a
an error bubbling up from a child atomic section. Most cases were
a try/catch block assumes that only the statement was rolled back
now result in an error and rollback.
Callers using try/catch to handle key conflicts should instead use
SELECT FOR UPDATE to find conflicts beforehand, or use IGNORE, or
the upsert()/replace() methods. The try/catch pattern is unsafe and
no longer allowed, except for some common errors known to just
rollback the statement. Even then, such statements can come from
child atomic sections, so committing would be unsafe. Luckily, in
such cases, there will be a mismatch detected on endAtomic() or a
dangling section detected in close(), resulting in rollback.
Remove caching from DatabaseMyslBase::getServerVariableSettings
in case some SET query changes the values.
Bug: T189999
Change-Id: I532bc5201681a915d0c8aa7a3b1c143b040b142e
* Refactor the code to be less cluttered, moving some logic
to reconnect(), which is now named replaceLostConnection()
* Handle the case of a second connection loss on query retry
* Make canRecoverFromDisconnect() check for temporary tables
* Do not clear session-level variables on deadlock errors
since only the transaction is lost, not the whole session
* Make sure any empty transaction is destroyed on deadlock
* Attempt reconnection *before* triggering the transaction
callbacks since some of them might want to use the database
* Define wasConnectionError() for postgres
* Remove unused return value from handleSessionLoss()
Change-Id: Ic1dcab03f087ce310637210e8e9bc0771e44f045
If the $var argument is provided, then it will make the resulting
count exclude rows where the value for that column is NULL.
Also add buildSelectSubquery() method and Subquery
wrapper class for use with select() for calculated tables.
Change-Id: I549d629af99afdf370602de095f7fba6d1546c37
Named locks are session-level constructs and this transaction agnostic.
Also make lockIsFree() a bit more consistent when the thread has the
lock itself.
Change-Id: Ief51196161bbc50c798740f3c738fd0e39880508
Done using the PhpStorm refactor->rename tool.
Also move "defaultBigSelects" declaration to DatabaseMysqlBase
as no other classes uses that.
Change-Id: I424a2d9815de3a5d4cca2522f3db23a5efe6b592
* Update replace()/upsert() to combine the affected row
count for the non-native case
* Also make replace() atomic in the non-native case,
similar to how upsert() already works
Change-Id: I6c9bcba54eca6bcf4a93a9b230aaedf7f36aa877
SQL supports parentheses for grouping in the FROM clause.[1] This is
useful when you want to left-join against a join of other tables.
For example, say you have tables 'a', 'b', and 'c'. You want all rows
from 'a', along with rows from 'b' + 'c' only where both of those
exist.
SELECT * FROM a LEFT JOIN b ON (a_b = b_id) JOIN c ON (b_c = c_id)
doesn't work, it'll only give you the rows where 'c' exists.
SELECT * FROM a LEFT JOIN b ON (a_b = b_id) LEFT JOIN c ON (b_c = c_id)
doesn't work either, it'll give you rows from 'b' without a
corresponding row in 'c'. What you need to do is
SELECT * FROM a LEFT JOIN (b JOIN c ON (b_c = c_id)) ON (a_b = b_id)
This patch implements this by extending the syntax for the $table
parameter to IDatabase::select(). When passing an array of tables, if a
value in the array is itself an array that is interpreted as a request
for a parenthesized join. To produce the example above, you'd do
something like
$db->select(
[ 'a', 'nest' => [ 'b', 'c' ] ],
'*',
[],
__METHOD__,
[],
[
'c' => [ 'JOIN', 'b_c = c_id ],
'nest' => [ 'LEFT JOIN', 'a_b = b_id' ],
]
);
[1]: In standards as far back as SQL-1992 (I couldn't find an earlier
version), and it seems to be supported by at least MySQL 5.6, MariaDB
10.1.28, PostgreSQL 9.3, PostgreSQL 10.0, Oracle 11g R2, SQLite 3.20.1,
and MSSQL 2014 (from local testing and sqlfiddle.com).
Change-Id: I1e0a77381e06d885650a94f53847fb82f01c2694
Under some conditions (Semantic MediaWiki, Gadgets), an integer is
passed to DatabaseMysqli::mysqlRealEscapeString (). This integer is, in
turn, passed to mysqli::real_escape_string (), which needs a string.
Under HHVM 3.19.1 (at least) this type mismatch causes an exception.
A typecast should prevent it.
I repeated the patch in other DB drivers where I could find a function
that escaped strings for SQL.
Bug: T163646
Change-Id: I1b7820bc064dc79498cf9f17747f745990c526b7
I0e6a9e6d overlooked the special handling PG needs (prior to 9.5 anyway)
to properly emulate MySQL's IGNORE option when delegating to the parent
implementation.
For now, then, don't use the native implementation in PG when IGNORE is
specified. Instead, fall back to the non-native implementation that does
a select() then an insert() where PG can handle the IGNORE properly.
In the future we might use the ON CONFLICT DO NOTHING clause added in PG
9.5 to be able to do native insertSelect() with IGNORE (and to better
handle multi-row insert() with IGNORE, and we could use the related ON
CONFLICT DO UPDATE to implement upsert()). All that is left for a future
patch.
Change-Id: I7987d59580543de03d5c6a5ed7fa3ce551ac12f3
It's often forgotten because MySQL and Sqlite don't use it, the only
users are PostgreSQL and Oracle. And when used, if inserts to multiple
tables are being done it's easy to get the ordering wrong.
This patch reimplements DatabasePostgres::insertId() in terms of PG's
lastval() function, and adds triggers to the Oracle schema to make it
work the same as the other databases.
Bug: T164900
Change-Id: Ib308190c52673a9266c8495a589ae644f9fbefce
* Fix schema for image_comment_temp.
* Provide values in CommentStoreTest::provideInsertRoundTrip() for
columns where the PG schema doesn't have a default value but the MySQL
schema does.
* Call nextSequenceValue() from CommentStoreTest::testInsertRoundTrip().
* Correctly handle $options being the string 'FOR UPDATE' in
DatabasePostgres::selectSQLText()
* Correctly handle the initial table in DatabasePostgres::selectSQLText() FOR
UPDATE mangling.
* Correctly handle aliases in DatabasePostgres::selectSQLText() FOR
UPDATE mangling.
Tests in PG are still going to be broken thanks to the fact that
nextSequenceValue() and insertId() can't be separated by another
nextSequenceValue()/insertId() pair. That should be taken care of by
T164898/T164900.
Change-Id: Ia770a003ca9170ab8bcc1436d8afe30700e00ada
selectField() and selectFieldValues() are trivial, they just need to
pass it through to select(). In fact, selectFieldValues() was already
doing it, just no one ever updated IDatabase.
insertSelect() is a little more work. nativeInsertSelect() was
originally written as largely a copy-paste of select() and has since
gotten well out of sync. Now that we have selectSQLText(), we should be
able to just use that. DatabasePostgres's implementation can wrap the
parent implementation instead of being another copy-paste, but
DatabaseOracle seems to still need to be special.
Change-Id: I0e6a9e6daa510639d3212641606047a5db96c500
The insertId() method was returning a string, which caused the
returnValueMap not to trigger due to int/string mismatches.
Also add sanity integer cast to WikiPage::insertOn().
Added a few more type docs.
Bug: T75174
Change-Id: Id1090f3e3d0481272a3d13c3af8f2588f06dc912
