laskoviymishka opened a new issue, #1052: URL: https://github.com/apache/iceberg-go/issues/1052
Parent: #589, follow-up to #996. The scanner DV read path is unit-tested at the loader level: `TestReadAllDeletionVectors` builds a puffin file in-process, drives `readAllDeletionVectors` directly, and asserts the returned `perFilePosDeletes` map shape. What's not yet covered is the end-to-end claim that DV-deleted row positions are actually absent from `GetRecords` output for a real Parquet data file. This is the most likely failure mode after a future refactor of the `processPositionalDeletes` pipeline: a change there would be type-correct (the compile-time signature assertion in `dv_scanner_read_test.go` pins the type) but semantically wrong, and no current test would catch it. Scope: write a small Parquet file with a handful of rows, pair it with a hand-built DV puffin blob marking a subset of positions deleted, construct enough v3 table metadata to drive a scan via the public `TableScan` API, then assert the returned Arrow record batches contain exactly the surviving rows. The fixture scaffolding overlaps with what a future cross-client DV write/read test would need, so this can be folded into that work when it materializes. Gated on #996 merging to main. -- 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]
