fallintoplace opened a new pull request, #1898: URL: https://github.com/apache/iceberg-go/pull/1898
## Summary - Read position deletes with `GetRecords` using only `file_path` and `pos`. - Process record batches incrementally with one grouping accumulator. - Avoid full-table materialization and `UnifyTableDicts`. - Keep the existing string, dictionary, validation, and ordering behavior. ## Performance Local benchmark on an Apple M1 Pro with Go 1.26.3. Each case used a 100 ms benchmark run and included an unused string column in the Parquet file. | Rows | Before | After | | ---: | ---: | ---: | | 100K | 11.1 ms/op, 77.7 MB/op, 1,009 allocs/op | 4.1 ms/op, 12.3 MB/op, 612 allocs/op | | 1M | 93.8 ms/op, 613.0 MB/op, 4,851 allocs/op | 26.6 ms/op, 75.6 MB/op, 1,960 allocs/op | The exact numbers will vary by machine and storage backend, but the projected reader avoids decoding and retaining unused columns. ## Tests - `go test ./table/... -count=1` - `go test -race ./table -count=1` - `go vet ./table` - `go test ./table -run ^$ -bench ^BenchmarkReadDeletesProjected/(rows=100000|rows=1000000)/(before|after) -benchtime=100ms -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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
