ocb3916 opened a new pull request, #4441:
URL: https://github.com/apache/streampark/pull/4441
## What is the purpose of the change
Implements Phase 0.2 of the Scala removal plan (#4408).
Defines `FlinkTableJob` and `FlinkStreamTableJob` as the Java abstract base
classes that will
eventually replace the Scala `FlinkTableTrait` and `FlinkStreamTableTrait`
traits — for Table API
(batch) jobs and jobs that mix the DataStream API with the Table API,
respectively. Both classes
live in the existing `streampark-flink-shims-base` module and follow the same
`ready → handle → destroy` lifecycle established by `FlinkStreamingJob` in
sub-issue 0.1, so all
three "Job" base classes share a consistent shape.
## Brief change log
- Add `FlinkTableJob` abstract class with lifecycle: `ready → handle →
destroy`, exposing the
underlying `TableEnvironment` via `getTableEnv()` (composition) instead of
extending / mimicking
the interface like the legacy trait did
- Add `FlinkStreamTableJob` abstract class with the same lifecycle, exposing
both
`StreamExecutionEnvironment` (`getEnv()`) and `StreamTableEnvironment`
(`getTableEnv()`) via
composition
- Add `sql(String)` / `sql(String, Consumer<String>)` on both classes,
bridging to the existing
Scala `FlinkSqlExecutor.executeSql` so the public API surface stays
Scala-free
- Add `markConvertedToDataStream()` on `FlinkStreamTableJob`, replacing the
legacy trait's
automatic `Table → DataStream` conversion detection (no longer possible
under composition) with
an explicit call
- Place both classes under `org.apache.streampark.flink.core.javaapi` rather
than the
issue-specified `org.apache.streampark.flink.core.java` — see note below
- Add JUnit 5 tests for both classes verifying lifecycle order,
`getTableEnv()`/`getEnv()`
composition, SQL execution, and the `markConvertedToDataStream()` flag
## A note on the package name
The issue body specifies `org.apache.streampark.flink.core.java`. That name
breaks Scala
compilation of every sibling file in `streampark-flink-shims-base` that
references
`java.util`/`java.io`/`java.lang` — a nested package literally named `java`
shadows the real
`java.*` standard library for any Scala file sharing the enclosing package.
Reproduced both ways
in a clean build:
- `org.apache.streampark.flink.core.java` → `BUILD FAILURE`, 33 errors, all
`java.util`/`java.io`/
`java.lang` unresolved in unrelated sibling `.scala` files
- `org.apache.streampark.flink.core.javaapi` → `BUILD SUCCESS`, no other
files affected
Went with `javaapi` for this PR. Open to other naming suggestions if
`javaapi` isn't preferred.
## Design note: composition instead of full delegation
The legacy traits extend/mimic Flink's `TableEnvironment` /
`StreamTableEnvironment` directly,
re-implementing ~40-50 delegate methods each (`FlinkStreamTableTrait`
additionally needed a
`$`-prefix naming hack to avoid clashes between the two APIs). The new
classes hold the
environment(s) as fields instead, exposed via `getTableEnv()` / `getEnv()`.
This removes the
delegate boilerplate and the naming hack, and means the classes won't go
stale as the Flink Table
API changes — but it **is a breaking change** for code that currently calls
Table API methods
directly on a trait instance (e.g. `job.sqlQuery(...)` →
`job.getTableEnv().sqlQuery(...)`).
Flagging for discussion since the Phase 11 migration guide will need to
cover it.
## Verifying this change
```bash
mvn test -pl
streampark-flink/streampark-flink-shims/streampark-flink-shims-base
```
Tests run: 9, Failures: 0, Errors: 0
## Does this pull request potentially affect one of the following parts?
- [x] The public API
- [ ] The web application
- [ ] The build system
## Documentation
- Relates to issue #4408
- Part of umbrella issue #4408, sub-issue 0.2
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
