[SkipRecovery] Restore randomized regex choice coverage [reduced-it] [databricks]

[wjxiz1992]

JaCoCo production line coverage: +0 lines (measured: delta-lake, iceberg, shuffle-plugin, sql-plugin, udf-compiler; Scala 2.12, shim 350, anchor 2d438c02, vs nightly coverage b34)

Fixes #10641.

Description

This test-only recovery restores randomized coverage for regex alternatives containing end-of-line anchors. test_regexp_choice still pins seed 0 after a historical CRLF capture mismatch, although the underlying fixes have merged. Production behavior is unchanged; the original failing input and seed now pass CPU/GPU comparison on current main.

Remove the temporary seed override and add the original aab2ab\r\n input to the existing generator's special cases. All 12 SQL expressions and their parity assertions remain unchanged. Other regex limitations and fallback tests are outside this change.

The underlying fixes are cuDF #22763 (CRLF anchor semantics) and #15023 (removal of the consuming CRLF rewrite).

AI assistance: The change and PR description were prepared with Codex assistance.

Validation

Apache Spark 3.5.0 / Scala 2.12 / Python 3.10.18 / Java 17 / UTC:

• Exact historical input: CPU and GPU both return ab2ab, without the trailing CR; GpuRegExpExtract presence was asserted.
• Complete regexp_test.py, seed 1711561563, xfail masking disabled: 95 passed, 1 skipped. The unchanged skip requires Spark 4.0+.
• Focused seeds 0, 20260911, and 42: 1 passed each. Original seed with OOM injection forced: 1 passed.
• Current-main shim-350 package build: BUILD SUCCESS.

[reduced-it] retains the affected test's only collected case: it has no parametrization decorators, and pre-commit reduced collection was checked. Scala UT scheduling remains parallel. Databricks validation is requested through the title marker; no local Databricks pass is claimed.

Reproduction commands and artifact evidence

Build source: e56c2660c762033f63e39df59c9276d5eb6e81e9.

mvn package -pl dist,integration_tests -am -Dbuildver=350 -DskipTests \
  -Dmaven.repo.local=./.mvn-repo \
  -Drapids.test.gpu.allocFraction=0.3 \
  -Drapids.test.gpu.maxAllocFraction=0.3 \
  -Drapids.test.gpu.minAllocFraction=0 \
  -s jenkins/settings.xml -P mirror-apache-to-urm

# With SPARK_HOME and both PYSPARK interpreters set to Spark 3.5 / Python 3.10:
TESTS=regexp_test.py DATAGEN_SEED=1711561563 TEST_PARALLEL=2 \
  integration_tests/run_pyspark_from_build.sh --runxfail --test_oom_injection_mode never

# Repeat with DATAGEN_SEED=0, 20260911, and 42:
TESTS=regexp_test.py DATAGEN_SEED=0 TEST_PARALLEL=1 \
  integration_tests/run_pyspark_from_build.sh -k test_regexp_choice \
  --runxfail --test_oom_injection_mode never

TESTS=regexp_test.py DATAGEN_SEED=1711561563 TEST_PARALLEL=1 \
  integration_tests/run_pyspark_from_build.sh -k test_regexp_choice \
  --runxfail --test_oom_injection_mode always

Built plugin SHA-256: 4b460b3390641dc617ccc2769f160e4824fbdde182c12d8ddb42715ce47d6d84. Its JNI dependency embeds cuDF revision f76d700abdd316b406ca825a269b73153632742d, which contains the merged CRLF fix.

Coverage: replayed the modified test successfully against the exact nightly class-anchor JAR (2d438c02049635368d2db1a7189b2093d206bad9, SHA-256 e38843bef80d8a33c33a6ba49344f11764cab64a3875747c866176fd5d6402fc) from coverage b34. Each module's regenerated baseline counters equal the published report, and merging the test execution adds no lines. No class-ID mismatch or MemoryCheckerImpl artifact was present. This restores randomized regression coverage, not new production-line coverage. The newer b21 report was not used because its Iceberg execution data has class-ID mismatches.

Checklists

Documentation

• Updated for new or modified user-facing features or behaviors
• No user-facing change

Testing

• Added or modified tests to cover new code paths
• Covered by existing tests
  (Please provide the names of the existing tests in the PR description.)
• Not required

Performance

• Tests ran and results are added in the PR description
• Issue filed with a link in the PR description
• Not required

[Copilot]

🟢 Approval recommended

Test-only changes are validated, with no unresolved review comments.

Pull request overview

Restores randomized regex-choice regression coverage for CRLF-anchored alternatives without production changes.

Changes:

• Removes the temporary fixed seed override.
• Adds the historical aab2ab\r\n input as a generator special case.

File summaries

File	Description
integration_tests/src/main/python/regexp_test.py	Updates regex test data generation.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[greptile-apps]

Confidence Score: 5/5

The test-only change appears safe to merge.

Summary

• Removes the permanent seed-0 override from test_regexp_choice.
• Adds aab2ab\r\n as a generated special case.
• Leaves all regex expressions and CPU/GPU parity assertions unchanged.

_{Reviews (1) · Last reviewed commit: "Restore randomized regex choice coverage..."}

[wjxiz1992]

build