github-actions[bot] commented on code in PR #69804:
URL: https://github.com/apache/doris/pull/69804#discussion_r4408536102
##########
be/src/format_v2/table_reader.cpp:
##########
@@ -1125,6 +1125,7 @@
TableReader::_find_table_column_identity_path_by_field_id(int32_t field_id,
Status TableReader::init(TableReadOptions&& options) {
_scanner_profile = options.scanner_profile;
+ _parquet_profile_cache = options.parquet_profile_cache;
Review Comment:
[P2] Forward the cache through HudiHybridReader children
FileScannerV2 now supplies this cache to the top-level HudiHybridReader, but
its `_init_child_reader` (be/src/format_v2/table/hudi_reader.cpp:194-211) omits
`.parquet_profile_cache = _parquet_profile_cache` when building the native/JNI
child options. The native Hudi child therefore keeps a null cache and every
Parquet file repeats ParquetProfile counter registration under the shared
RuntimeProfile counter-map lock, so Hudi MOR scans with many native files still
pay the per-file setup this change is intended to remove. Please forward the
field here as Paimon does and add hybrid native-split coverage.
--
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]
