lukecwik commented on a change in pull request #12639:
URL: https://github.com/apache/beam/pull/12639#discussion_r474144618
##########
File path:
sdks/java/harness/src/test/java/org/apache/beam/fn/harness/BeamFnDataReadRunnerTest.java
##########
@@ -352,6 +354,35 @@ public void testRegistration() {
fail("Expected registrar not found.");
}
+ @Test
+ public void testSplittingBeforeStartBundle() throws Exception {
+ List<WindowedValue<String>> outputValues = new ArrayList<>();
+ BeamFnDataReadRunner<String> readRunner =
+ createReadRunner(outputValues::add, PTRANSFORM_ID,
mockBeamFnDataClient);
+ // The split should happen at 5 since the allowedSplitPoints is empty.
+ assertEquals(
Review comment:
A bundle can be started before each transforms `start` method has been
invoked since we invoke `start` starting from the leaves and going to the
parents and then we invoke `finish` in reverse order.
Having this definition of `started` allows us to get msec counters for
`start` methods since they can be slow, checkpointing without processing
through this read transform...
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
