Details
-
Bug
-
Status: Resolved
-
P2
-
Resolution: Fixed
-
None
Description
Beam translates Combine.Globally as a composite transform composed of:
- Map that assigns Void keys
- Combine.PerKey
on spark: As Combine.Perkey uses a spark GBK inside it, the runner adds its own translation of Combine.Globally to avoid less performant GBK. This translation should be called in place of entering the composite transform translation.A pipeline like this:
PCollection<Integer> input = pipeline.apply(Create.of(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)); input.apply( Combine.globally(new IntegerCombineFn()));
private static class IntegerCombineFn extends Combine.CombineFn<Integer, Integer, Integer> { @Override public Integer createAccumulator() { return 0; } @Override public Integer addInput(Integer accumulator, Integer input) { return accumulator + input; } @Override public Integer mergeAccumulators(Iterable<Integer> accumulators) { Integer result = 0; for (Integer value : accumulators) { result += value; } return result; } @Override public Integer extractOutput(Integer accumulator) { return accumulator; } }
is translated as the above composite.