linliu-code opened a new pull request, #18689:
URL: https://github.com/apache/hudi/pull/18689

   ### Describe the issue this Pull Request addresses
   
   Follow-up to #18224 (JsonKinesisSource). Small correctness fixes that 
surfaced after the initial merge.
   
   ### Summary and Changelog
   
   1. **Numeric sequence-number comparison.** Kinesis sequence numbers are 
128-bit integers represented as decimal strings whose lengths can vary, so 
`String.compareTo` is not correct for ordering them. Introduce 
`KinesisOffsetGen.CheckpointUtils.compareSequenceNumbers(a, b)` using 
`BigInteger`, and use it in:
      - `KinesisShardRange.hasUnreadRecords` (closed-shard fully-consumed check)
      - `KinesisOffsetGen.checkDataLossOnExpiredShards` (also adds an 
empty-`lastSeq` guard for `|endSeq` checkpoints — a closed shard that has not 
yet had any record consumed is treated as not consumed)
      - `KinesisOffsetGen.checkDataLossOnAvailableShards` (trim-horizon check)
   
   2. **Null-safe `millisBehindLatest`.** 
`GetRecordsResponse#millisBehindLatest()` is documented as nullable in the AWS 
SDK; the previous `response.millisBehindLatest() == 0` would NPE on auto-unbox 
if a server returned null. Extract to a `Long` and null-check before comparing.
   
   3. **Integer-overflow guard on per-call limit.** `(int) (maxTotalRecords - 
totalConsumed)` could silently truncate if the remaining budget exceeded 
`Integer.MAX_VALUE`. Switch to `Math.toIntExact(...)` so it fails loudly 
instead.
   
   4. **Explicit getter names on `KinesisReadConfig`.** Replace the 
Lombok-generated `isShouldAddMetaFields()` / `isEnableDeaggregation()` with 
explicit `shouldAddMetaFields()` / `isDeaggregationEnabled()` (kept under 
`@Getter` with `AccessLevel.NONE` on the two boolean fields). Updates the two 
call sites in `JsonKinesisSource`.
   
   ### Impact
   
   No user-visible behavior change in the common case. The numeric-comparison 
fix only matters when comparing two Kinesis sequence numbers of different 
decimal lengths; \`Math.toIntExact\` and the null guard are defensive and only 
change behavior on edge-case inputs that previously truncated or NPE'd.
   
   ### Risk Level
   
   Low
   
   ### Documentation Update
   
   none
   
   ### Contributor's checklist
   
   - [x] Read through [contributor's 
guide](https://hudi.apache.org/contribute/how-to-contribute)
   - [x] Enough context is provided in the sections above
   - [x] Adequate tests were added if applicable — existing Kinesis tests 
(TestKinesisCheckpointUtils, TestJsonKinesisSource, TestShardRecordIterator, 
TestKinesisShardRange, TestKinesisSourceFiltering, TestKinesisDataLossChecks) 
all pass: 80 tests, 0 failures, 0 errors, 2 pre-existing skips


-- 
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