zjw1111 opened a new pull request, #510:
URL: https://github.com/apache/paimon-cpp/pull/510

   ### Purpose
   
   Linked issue: none.
   
   Fix two Hive Decimal hashing differences from Java: include every 32-bit 
word of the unscaled magnitude in the hash, and apply HALF_UP once when 
reducing scale by multiple digits. For example, Decimal 4294967296 now hashes 
to 961, and unscaled=149, scale=40 hashes to 69.
   
   Add focused coverage for GenericArray, default file-index predicates, 
histogram windows, object-store filesystem operations, REST message 
serialization, and Parquet filtering/fallback paths. The Parquet tests cover 
missing page indexes, disjoint predicate/bitmap ranges, no-predicate index I/O, 
and recovery from a synchronous prebuffer exception. Decimal overflow tests 
document the existing C++ zero fallback and Java null/NullPointerException 
behavior.
   
   ### Tests
   
   Debug/ASan validation:
   - `cmake --build build --target paimon-common-test paimon-core-test 
paimon-parquet-format-test -j 64`
   - `./build/debug/paimon-common-test`: 1807 passed, 2 skipped.
   - `./build/debug/paimon-core-test 
--gtest_filter='HiveHasherTest.*:HiveBucketFunctionTest.*'`: 26 passed.
   - `./build/debug/paimon-parquet-format-test 
--gtest_filter='FileReaderWrapperTest.*:*ParquetFileBatchReaderTest.*'`: 62 
passed.
   - REST message tests: 22 passed in an existing standalone ASan build.
   - Object-store filesystem/input-stream tests: 17 passed in an existing 
standalone ASan build.
   
   The local main build disables REST, OSS, and S3, so REST and object-store 
tests were compiled separately against the local project libraries, using `-j 
64`.
   
   All changed files pass `pre-commit run --files ...` and `git diff --check`.
   
   An earlier full core run passed 2545 tests and failed the two 
`WithOptionalIOManager/MergeTreeWriterTest.TestSortedReaderFailure` variants: 
exact Status comparisons differ when `PAIMON_EXTRA_ERROR_CONTEXT` adds 
call-site context. Those paths are outside this change.
   
   ### API and Format
   
   No public API, file format, or protocol changes. Corrected Decimal hashes 
can change Hive bucket assignments for affected values to match Java.
   
   ### Documentation
   
   Inline comments explain the Decimal hash calculation and Java overflow 
behavior. No new user-facing feature documentation is needed.
   
   ### Generative AI tooling
   
   Generated-by: OpenAI Codex 0.163.0
   


-- 
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]

Reply via email to