Conversation
Also close the reader at end of reader. Updated the test for reader reuse.
[BEAM-3552] Support Python 3 in the metrics, internal, typehints, and utils modules.
…imers for the DirectRunner. [BEAM-3153] Add processing-time timers for the DirectRunner.
to use of `extractOrderedList()` which removes all the elements from the heap. `extractOrderedList()` is costly and is not required either. `extractOutput()` does not mutate now and is cheaper too. Updated LargestUnique.add() to avoid 'heap.contains()' call in common case with large input. Merging was not tested as direct-runner does not seem to use combiner. Added test using ConbineFnTester. Put back add() improvement. contains() is an O(n) operation. Avoid it in common case. I think the extractOrderedList() existed mainly to avoid this.
…unner [BEAM-3689] Fix unbounded reader leak in direct-runner.
…to enforce blank lines for imports and classdefs
Print correct line numbers for warnings.
Add missing import statements.
…eout [BEAM-3775] Increase timeout in beam_PostCommit_Java_ValidatesRunner_…
….0-SNAPSHOT Bump sdks/go/container/pom.xml to 2.5.0-SNAPSHOT
[BEAM-3791] Update version number in build_rules.gradle
…on all runners use to simplify the contract the user has to implement
…form names on Flink operators
… state method defaulting to the implementation all runners use to simplify the contract the user has to implement [BEAM-3794] Make StateInternals short state method defaulting to the implementation all runners use to simplify the contract the user has to implement
This re-instantiates the Go SDK. The resulting situation is this:
-----A--A*----------------- [master]
/ \ \ \
/ A** \ \
/ \ \ \
-----*--*---*-------------- [go-sdk]
A accidental merge of Go SDK to master
A* revert of A; removes the Go SDK from any branch it is merged to
A** revert of A* to put Go SDK back on the go-sdk branch
This makes all expected upstream merges and eventual
promotion of the Go SDK work as expected.
lukecwik
left a comment
There was a problem hiding this comment.
I went through the same git steps to revert and re-revert to validate the change and looked at the merge tree.
|
Note that #4808 is the change to fix up the master branch. |
|
I'll wait for #4808 to go green and get merged to |
|
I've validated the Go build still works, so LGTM from that perspective. |
|
Were the RAT failures pre-existing? |
|
It turns out that the Go/Java precommits don't depend on running RAT so it is being skipped: Line 294 in 9eca13c |
|
Why is the precommit looped out and using those targets? Shouldn't it just be straight line hardcoded list of targets like |
|
Ken, please direct your comments on #4830 |
Into the Go SDK branch, merges
masterup to revert of accidental Go SDK, plus re-revert to put the Go SDK back.