github-actions[bot] commented on code in PR #66802:
URL: https://github.com/apache/doris/pull/66802#discussion_r3793662130
##########
be/src/format_v2/file_reader.h:
##########
@@ -347,6 +357,19 @@ class FileReader {
// Initialize file reader and parse file metadata.
virtual Status init(RuntimeState* state);
+ // Optionally refine one scheduler split into format-specific physical
children after metadata
+ // initialization. The default keeps non-columnar formats on their
original split.
+ virtual Status build_physical_splits(const FileScanSplit& source_split,
Review Comment:
[P1] Keep split policy above FileReader
Please keep this interface format-local. `FileScanSplit` contains
scanner/table policy—FE source identity, source progress/reservation,
`clear_table_level_row_count`, and cross-child Condition Cache coordination—and
`ParquetReader::build_physical_splits()` now constructs those fields below
`TableReader`. This directly conflicts with the required Format V2 boundary:
`FileReader` may expose physical metadata but must not own Scanner scheduling,
Split-source, or table-format semantics. Return a neutral descriptor such as
`{format_split_id, byte_range, shared_file_context}` here, then have
`TableReader`/`FileScannerV2` wrap it in `FileScanSplit` and attach those
policies.
--
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]
