fallintoplace opened a new pull request, #1216: URL: https://github.com/apache/arrow-go/pull/1216
## Summary - **Batch sparse union appends** through the child bulk APIs. - **Write repeated type IDs and dense offsets** directly into their buffers. - **Keep dense unions compact** by sharing one physical null or empty child value. - **Treat non-positive counts as no-ops** for both union modes. - Add parity, edge-case, and benchmark coverage. ## Benchmarks Apple M1 Pro, darwin/arm64, 65,536 rows. Times are rounded medians from 3 runs. | Mode | Operation | Before | After | Before allocs | After allocs | | --- | --- | ---: | ---: | ---: | ---: | | Sparse | nulls | ~1.83 ms | ~0.35 ms | 44 | 34 | | Sparse | empty | ~1.83 ms | ~0.33 ms | 44 | 34 | | Dense | nulls | ~0.55 ms | ~0.14 ms | 51 | 29 | | Dense | empty | ~0.39 ms | ~0.14 ms | ~51 | 29 | The benchmark uses the same builder setup before and after. Small row counts stay close to benchmark noise. ## Tests - `go test ./arrow/array -count=1` - `go test -race ./arrow/array -run 'Test(UnionBuilderBulkAppendNullsAndEmptyValues|UnionBuilderZeroBulkAppendDoesNotMutateChildren|UnionBuilderBulkAppendMatchesScalar)$' -count=1` - `go vet ./arrow/array` - `PARQUET_TEST_DATA="$PWD/parquet-testing/data" go test -count=1 ./...` -- 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]
