* destination-specification: add supportsNormalization and supportsDBT attributes
* address review comment
* missed this one
* output after gradle format
* destination: add implementation for mysql as destination
* Fix formatting errors.
* address review comments + fix flaky test
* fix formatting
* address Davin's review comments
* add missing todo
* enable namespace test + only provide test user the minimum permissions required
Co-authored-by: Davin Chia <davinchia@gmail.com>
This PR introduces the following behavior for JDBC sources:
Instead of streamName = schema.tableName, this is now streamName = tableName and namespace = schema. This means that, when replicating from these sources, data will be replicated into a form matching the source. e.g. public.users (postgres source) -> public.users (postgres destination) instead of current behaviour of public.public_users. Since MySQL does not have schemas, the MySQL source uses the database as it's namespace.
To do so:
- Make namespace a field class concept in Airbyte Protocol. This allows the source to propagate namespace and destinations to write to a source-defined namespace. Also sets us up for future namespace related configurability.
- Add an optional namespace field to the AirbyteRecordMessage. This field will be set by sources that support namespace.
- Introduce AirbyteStreamNameNamespacePair as a type-safe manner of identifying streams throughout our code base.
- Modify base_normalisation to better support source defined namespace, specifically allowing normalisation of tables with the same name to different schemas.
Last step (besides documentation) of namespace changes. This is a follow up to #2767 .
After this change, the following JDBC sources will change their behaviour to the behaviour described in the above document.
Namely, instead of streamName = schema.tableName, this will become streamName = tableName and namespace = schema. This means that, when replicating from these sources, data will be replicated into a form matching the source. e.g. public.users (postgres source) -> public.users (postgres destination) instead of current behaviour of public.public_users. Since MySQL does not have schemas, the MySQL source uses the database as it's namespace.
I cleaned up some bits of the CatalogHelpers. This affected the destinations, so I'm also running the destination tests.
This PR is step 5 of this tech spec - https://docs.google.com/document/d/1qFk4YqnwxE4MCGeJ9M2scGOYej6JnDy9A0zbICP_zjI/edit.
The first of (at least) 2 PRs to implement this on the source side. I made some headway before deciding to break the changes into one PR implementing this for discover schema job, and another PR implementing this for read. The combined PR would have been too big otherwise.
Also refactor MoreResources as the test method was attempting to write to the location classes where loaded out from - the issue is we cannot guarantee where the class is loaded from can be written to. Changing this to write to a random folder in the temp directory.
Add namespace field to the Airbyte Stream in preparation to propagate a source defined namespace to the Destination.
This namespace field is then consumed as the destination schema the table is written to.
This only applies to JDBC destinations.
This is Steps 1 - 4 of the namespace tech spec, seen at https://docs.google.com/document/d/1qFk4YqnwxE4MCGeJ9M2scGOYej6JnDy9A0zbICP_zjI/edit.
Some minor refactoring and commenting as I go.
* Remove unnecessary test classes as they match Integration tests in terms of what is being tested. They have no real value since the corresponding integration test can be run locally without additional credentials. The main value the classes brings is letting us run tests without building the docker image (the integration tests require doing so). however I feel this benefit is not worth the additional maintenance cost.
* Centralise DataArgumentProvider into it's own class for easier maintenance and usability.
* Handle destination sync mode in destinations
* Source & Destination sync modes are required (#2500)
* Provide Migration script making sure it is always defined for previous sync configs
* Add standard tests for sources that use the JdbcSource to guarantee that changes do not break any sources that rely on JdbcSource.
* Add JdbcStressTest to verify that we stream / chunk data properly (a.k.a can handle more data in any JdbcSource than fits in memory)
* Migrate MSSQL and Redshift to user the new base source
* add stream and field name validation in discovery
* fix resource handling
* filter out streams with invalid stream or field names in the sync worker
* remove accidental commit
* add unicode key to test
* add another unicode test to isValidIdentifier
* update docs to reflect this change
* support using jsonschema types within sources + spec/read support for exchangeratesapi
* remove unnecessary key sorting
* projectDir/../.. -> rootDir
* better log and record handling
* output state messages and exclude outputting schema as message
* fix test
* hide warnings
* fmt
