leaves12138 opened a new pull request, #9908: URL: https://github.com/apache/paimon/pull/9908
### Purpose Extract the Python sidecar core and read integration from #9743, on top of the merged Java v1 format and read support (#9845, #9886, #9889, #9897). ### Changes - Add the standalone v1 sidecar codec, complete block directory, exact row-ID intervals, partition dictionary and bucket coverage, using the shared Java golden file. - Prune manifest blocks before decoding entries in the Python scan path. Preserve normal entry predicates, column groups and ADD/DELETE reconciliation. - Read only explicitly referenced sidecars, fall back to normal manifest reads for missing/unusable sidecars, and propagate cancellation/interruption. - Skip sidecar I/O when disabled, without relevant filters, and in explain/statistics mode. - Classify sidecars as metadata so existing local FileIO caching can reuse their bytes. ### Scope This is a **read/core-only draft**. The low-level codec can produce bytes for fixtures and callers, but Python manifest writers do not automatically create or publish sidecars in this PR. Writer, merger and failed-write cleanup changes are deferred to a follow-up. No Java cache changes are included. ### Validation - Run on dev2 from a clean checkout based on current master. - Python 3.11: 159 manifest/read/file-type tests passed. - Python 3.6: all 91 focused sidecar/file-type tests passed with isolated dependencies. The broader suite has 133 passed, 9 skipped and 17 failures; the same 17 failures reproduce on unmodified master in the same environment (PyArrow 6 lacks `Table.from_pylist` / `sort_by`, and Lance is not installed). - Shared Java v1 golden file: byte-for-byte encoding agreement and read selection checks passed. - Python lint passed using the repository configuration. - AST comparison confirms only `read` and `read_entries_parallel` changed in `ManifestFileManager`; all writer methods remain identical to master. -- 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]
