* chore: increase oracledb version
* feat: optionally use thick client
BREAKING CHANGE: With node-oracledb the thin client is used as default. Added a option to use the thick client. Also added the option to specify the instant client lib
closes: #10277
* fix: make thick driver the default, to assure an easy transition from version 5 to 6
* fix: make thin driver default again
Since there should be no problems with oracle db version >= 12.1
For oracle db < 12.1 the thick client must be used.
* chore: increase oracledb to 6.1.0
* chore: increase oracledb to 6.2.0 and added new option binaryDir
* refactor: fix linting
* update oracledb to 6.3.0
* feat: add support for mssql@10
Add mssql@10 to peerDependency, alongside mssql@9.
Update tests to use latest version of mssql and Node v16.
Closes: #10340
* fixed failing test
---------
Co-authored-by: Alex Messer <dmzt08@gmail.com>
* fix: extend spatial column types for postgres driver
Extend spatial column type with range types for postgres. This fixes a bug when using spatial indices.
Closes: #10567
* format
* remove only from test, ready for pr
* changed a way how we determine index type
---------
Co-authored-by: Dmitry Zotov <dmzt08@gmail.com>
* Add a replication test suite
* Allow overriding the default destination for read queries when using replication
For many applications that want to use read slaves, data consistency isn't super important, and so all reads can be sent by default to a read slave. For some applications though, it's only select queries that can be sent to read slaves as end-users of the application expect data consistency almost everywhere. This new setting lets apps like those change the default such that query builders need to opt *in* to sending to a read slave, instead of opt out.
Between v3 and v4 in node-redis the client options changed for a collection of socket related
paramaters. In the caching documentation the v3 format for used for `host` and `socket`. When
following the documented example the redis client would default to `localhost` even when a
different value for `host` was specified.
Closes: #9987
* test: fix migration-generate test suite
WIP
Closes: #10360
* fixed mssql test hang;
added option to disable process exit;
---------
Co-authored-by: Alex Messer <dmzt08@gmail.com>
For previous examples typescript error occurs in createQueryBuilder that says “Argument of type 'string' is not assignable to parameter of type 'QueryRunner'.ts(2345)”.
If you check the type of createQueryBuilder, the string cannot come as the first argument.
The types of createQueryBuilder are as follows.
myDataSource.createQueryBuilder<ObjectLiteral>(entityClass: EntityTarget<ObjectLiteral>, alias: string, queryRunner?: QueryRunner | undefined):
Disabling the raft log is going to be disallowed in CockroachDB. Also,
this setting wasn't having any effect, since the tests use an in-memory
cluster.
* fix: ensure comparing EntityMapIds in `buildForSubjectRelation`
I have found the issue (for me at least). In the method `buildForSubjectRelation` of `ManyToManySubjectBuilder`, in order to know if it is needed to insert the relation rows `databaseRelatedEntityIds` is compared with `relatedEntityRelationIdMap` (line 154). The problem is that `databaseRelatedEntityIds` is not always an "EntityIdMap". If one is overloading the entity in the `afterLoad` event of an EventSubscriber for example the line 102 :
`const databaseRelatedEntityValue = relation.getEntityValue(subject.databaseEntity);` will return an "enhanced" object with additional properties. To fix this, in this commit, we ensure that `databaseRelatedEntityIds` is containing an "EntityIdMap".
Closes#5704
* refactor: describe.only -> describe
removes describe.only.
* style: npm run format
format code.
* Update issue-5704.ts
refactor: typo on post title
---------
Co-authored-by: blemoine <services+bitbucket@blemoine.pro>
