He-Pin opened a new pull request, #3684:
URL: https://github.com/apache/pekko/pull/3684
### Motivation
A command handler can select between persisting and replying, replying
without persistence, or explicitly not replying. After that decision, a caller
cannot append a common callback through the staged reply-effect API introduced
in #3682. Recording a metric or sending a notification requires duplicating the
callback in each branch or moving it before the reply.
Add `thenRun` to that stage, while making `thenStop` end callback
composition:
```scala
handleCommand(state, command)
.thenRun(updatedState => recordMetric(updatedState))
.thenStop()
```
The helper returns `ReplyEffectWithStop`. `thenStop().thenRun(...)` must not
compile.
### Modification
- Declare `thenRun` on `ReplyEffectOps`, returning its recursive `Self`
type: Scala's state callback and Java's state and no-argument overloads, for
both event-sourced and durable-state persistence. Mark all new methods `@since
2.0.0`.
- Change `ReplyEffectWithStop.thenStop()` to return the base `ReplyEffect`.
This rejects callbacks after stopping, including through `thenUnstashAll`,
while `thenRun(...).thenUnstashAll().thenStop()` retains the expected stage.
`stash()` continues to expose only the base type.
- Keep traits abstract and implement the Java overloads in `EffectImpl`.
Make the existing Java `EffectBuilder.thenRun` methods overridable while
retaining their bodies and descriptors. The existing subtype callback
adaptation is reused; no effect downcasts or runtime stage checks are
introduced.
- Add matching Scala and Java tests for persistence/reply/callback ordering,
common callbacks after accepted and rejected branches, both no-reply entry
points, generic `Self` helpers, and stopping after callbacks. Add compile-time
rejection coverage for stopped, stashed, and widened effects.
- Update both persistence guides with the use case, type restrictions, and
existing at-most-once side-effect guarantees. Sending a reply does not wait for
the recipient to receive or process it.
### Result
Callers can append callbacks after choosing a reply without losing the
ability to unstash or stop. Stopping the reply-effect chain removes callback
capabilities. This adds no new trait and changes no runtime side-effect
ordering; existing `EffectBuilder` chaining remains outside this reply-stage
restriction.
**Stacked on #3682 at `c00fbb3376d47b8ec28abe151673207ec5f4e71e`.** The base
branch `stack/3682-reply-effect-then-stop` mirrors that exact commit so this PR
shows only the incremental change and runs upstream CI. Keep this PR in draft
until the parent merges; then rebase and retarget to `main` before merging. Do
not merge into the temporary stack base.
### Tests
- Isolated Scala 2.13 compilation of the eight effect API/implementation
sources and all four new test files: passed.
- Scala and `javac` client probes: valid generic/fluent chains compile; 10
Scala and 20 Java invalid chains are rejected. Restoring the old `thenStop`
return type makes the stopped-chain probes compile, demonstrating the
regression check. Positive `thenRun` probes fail against the parent API as
expected.
- Direct MiMa 1.2.1 probe against compiled `origin/main` sources: the same
22 findings covered by the parent's exact filters; no additional filter added.
Published-baseline cross-version MiMa remains for CI.
- `scalafmt --mode diff-ref=c00fbb3376d47b8ec28abe151673207ec5f4e71e` and
`--list`: passed.
- `sbt headerCreateAll javafmtAll` (JDK 17), followed by `sbt
+headerCheckAll javafmtCheckAll checkCodeStyle`: passed.
- `sbt sortImports`: environment/tool linkage failure in `stream-tests` and
`multi-node-testkit` (`scala.meta.package$.Input()`); unrelated generated edits
removed. Subsequent code-style checks passed.
- `git diff --check`: passed. Auditable Qoder stdout review: no must-fix
findings.
- Full runtime tests, Scala 3/next, `validatePullRequest`,
published-baseline MiMa, and Paradox: delegated to remote CI; pending. No local
full-suite runtime result claimed.
### References
Refs #3682, Refs #3683
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
