PHOENIX-7267 CsvBulkLoadTool fails job due to a bad record with "(sta…

[xavifeds8]

…rtline 1) EOF reached before encapsulated token finished"

[xavifeds8]

With commons-csv 1.0, CsvBulkLoadTool would fail the entire MapReduce job when encountering a malformed CSV record.
After the upgrade of commons-csv to 1.14.1
--ignore-errors: Bad records are skipped, good records are loaded, errors are counted in MR counters
Without --ignore-errors: Job fails gracefully with a clear error message instead of crashing

Sanity test for the upgrade : https://gist.github.com/xavifeds8/bd6015a1733ddbf630cbbdb453bdbc0d

[xavifeds8]

Changes made :

1. Upgraded commons-csv from 1.0 to 1.14.1
2. Migrated deprecated CSVFormat.withXxx() calls to CSVFormat.Builder API
3. Migrated deprecated new CSVParser(reader, format) to CSVParser.builder().setFormat(format).setReader(reader).get()
4. Caught UncheckedIOException (thrown by commons-csv 1.14.1 during iteration) in UpsertExecutor and CsvToKeyValueMapper, so parse errors are now routed through the normal error-handling path
5. Updated Pherf's CSVFileResultHandler and GoogleChartGenerator for the same API migration

Tests:

1. Updated existing CSVCommonsLoaderIT tests for the new API
2. Fixed testCSVCommonsUpsertBadEncapsulatedControlChars assertion to match the new exception wrapping
3. Added testCSVCommonsUpsertEOFInEncapsulatedToken — directly tests the reported scenario (unclosed quote at EOF)

[xavifeds8]

Hi @virajjasani, could you please trigger a CI run for this PR? Also, let me know if the changes look good to you.

[virajjasani]

@xavifeds8 the CI builds are broken temporarily. Could you run all csv bulk load realted tests and some more in your local to confirm nothing is broken?

[virajjasani]

+1, need manual verification of build and some test results

[NihalJain]

Please ensure that the newer csvParser actually throws UncheckedIOException for a bad record. We should not catch anything more unless needed

[xavifeds8]

Hey @NihalJain
https://github.com/apache/commons-csv/blob/6f93c7edfa0f758f757227b1d30588411fdbf669/src/main/java/org/apache/commons/csv/CSVParser.java#L234

Here in 1.14.1 in csv-commons CSVParser IOException is wraped with UncheckedIOException.

also have added a UT to verify this behaviour 25be349

[INFO] -------------------------------------------------------
[INFO] T E S T S
[INFO] -------------------------------------------------------
[INFO] Running org.apache.phoenix.mapreduce.CsvToKeyValueMapperTest
[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 0.064 s -- in org.apache.phoenix.mapreduce.CsvToKeyValueMapperTest
[INFO]
[INFO] Results:
[INFO]
[INFO] Tests run: 1, Failures: 0, Errors: 0, Skipped: 0
[INFO]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 23.525 s
[INFO] Finished at: 2026-05-18T12:02:52+05:30
[INFO] ------------------------------------------------------------------------

% cat phoenix-core/target/surefire-reports/org.apache.phoenix.mapreduce.CsvToKeyValueMapperTest-output.txt

Exception type: java.io.UncheckedIOException
Message: org.apache.commons.csv.CSVException: (startline 1) EOF reached before encapsulated token finished
Cause type: org.apache.commons.csv.CSVException
Cause message: (startline 1) EOF reached before encapsulated token finished
phoenix %