Gabriel39 opened a new pull request, #65151: URL: https://github.com/apache/doris/pull/65151
## Summary Add best-effort Parquet column-chunk prefetch for the FileScannerV2 Parquet path. ## Changes - Expose `ParquetFileContext::prefetch_ranges()` to warm selected Parquet byte ranges through `CachedRemoteFileReader`. - Build row-group prefetch ranges from projected Parquet leaf column chunks, including nested projections. - Trigger predicate-column prefetch when opening a selected row group. - Trigger non-predicate column prefetch immediately for scans without row-level filters, or after predicates leave surviving rows for lazy-read scans. ## Notes This keeps the Arrow Parquet random-access reader unchanged. Prefetch is a file-cache warm-up only and becomes a no-op when the underlying reader is not `CachedRemoteFileReader`. ## Validation - `/usr/local/opt/llvm@16/bin/clang-format -i be/src/format_v2/parquet/parquet_file_context.h be/src/format_v2/parquet/parquet_file_context.cpp be/src/format_v2/parquet/parquet_scan.h be/src/format_v2/parquet/parquet_scan.cpp` - `git diff --check -- be/src/format_v2/parquet/parquet_file_context.cpp be/src/format_v2/parquet/parquet_file_context.h be/src/format_v2/parquet/parquet_scan.cpp be/src/format_v2/parquet/parquet_scan.h` BE UT was attempted with `./run-be-ut.sh --run --filter='FileScannerV2Test.*:ParquetPageCacheTest.*'`, but local configuration failed before compiling tests because `nproc` is missing and CMake could not find Snappy in `thirdparty/installed`. -- 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]
