felipepessoto opened a new pull request, #13135:
URL: https://github.com/apache/gluten/pull/13135
<!--
Thank you for submitting a pull request! Here are some tips:
1. For first-time contributors, please read our contributing guide:
https://github.com/apache/gluten/blob/main/CONTRIBUTING.md
2. If necessary, create a GitHub issue for discussion beforehand to avoid
duplicate work.
3. If the PR is specific to a single backend, include [VL] or [CH] in the PR
title to indicate the
Velox or ClickHouse backend, respectively.
4. If the PR is not ready for review, please mark it as a draft.
-->
## What changes are proposed in this pull request?
<!--
Provide a clear and concise description of the changes introduced in this PR.
Ensure the PR description aligns with the code changes, especially after
updates.
If applicable, include "Fixes #<GitHub_Issue_ID>" to automatically close the
corresponding issue
when the PR is merged.
-->
**Depends on #13133.** This branch includes its verified JNI
throwable-carrier commit (`7e977538f`), followed by the separate native-cast
commit (`cbae44487`, 13 files). Both head branches are in a fork, so this PR
targets upstream `main` and its upstream diff is temporarily cumulative until
the lower PR merges. [Review only the native-cast
layer](https://github.com/felipepessoto/gluten/compare/felipepessoto-jni-exception-preservation...felipepessoto-native-cast-exceptions).
No native GitHub stack registration is used for these fork heads.
Translate native-origin cast failures that currently arrive as generic
`GlutenException`s, targeting the eight scalar/struct-member shapes behind the
16 ANSI-enabled `ImplicitStreamingMergeCastingSuite` failures.
- At the Velox output boundary, recover original Java throwables first. Only
then recognize native `USER` / `INVALID_ARGUMENT` errors attributed to a
built-in `cast` by `ExpressionExceptionProperties`.
- Carry the separate native reason and diagnostic in a typed
`NativeCastException`, using UTF-8 byte arrays and the existing `JavaException`
transport from #13133. There is no string-envelope protocol or exception-class
reconstruction from stack traces.
- Extend `ColumnarBatchOutIterator.translateException` with a Spark-package
helper that calls real `QueryExecutionErrors` factories: numeric narrowing
becomes `SparkArithmeticException` / `CAST_OVERFLOW`; string-to-integral
failures become `SparkNumberFormatException` / `CAST_INVALID_INPUT`; decimal
precision failures use Spark's version-specific `NUMERIC_VALUE_OUT_OF_RANGE`
factory (`.WITH_SUGGESTION` on Spark 4).
- Preserve the native diagnostic as the cause and retain schema translation
and the existing fallback for unrecognized or ambiguous reasons. Recognition
uses a narrow, type-checked native reason grammar; embedded unescaped string
delimiters are rejected rather than guessed.
No changes to overflow evaluation, ANSI support, legacy/TRY cast behavior,
or ARRAY/MAP casts. Original-Java preservation is owned by #13133 rather than
reimplemented here. Developer documentation explains the native-cast boundary
and grammar limitation.
## How was this patch tested?
<!--
Describe how the changes were tested, if applicable.
Include new tests to validate the functionality, if necessary.
For UI-related changes, attach screenshots to demonstrate the updates.
-->
**Executed locally:**
- Compiled the actual new Java carrier, Scala helper, and checked-in
`GlutenNativeCastExceptionSuite` against real Spark APIs in an isolated JVM
harness, resolving dependencies through `build/mvn`. All **13 tests passed on
each of Spark 3.4.4, 3.5.5, 4.0.2, and 4.1.1** (52 passing cases; Scala
2.12/2.13, JDK 17). These assert exact Spark exception classes, error classes,
message parameters, SQLSTATE, original-cause identity, UTF-8/NUL handling, and
unknown/malformed/ambiguous negative controls. No mock Spark exception
implementations were used.
- Verified the compiled JNI constructor descriptor `([B[B)V` and
Java-callable Spark helper descriptor with `javap`.
- `dev/format-cpp-code.sh` and targeted `clang-format-15 --dry-run
--Werror`, using clang-format 15.0.7.
- `dev/format-scala-code.sh` and its `check` mode, scoped to changed
Java/Scala files with `spotlessFiles`.
- The actual CI license-header fixer/checker,
`.github/workflows/util/license-header.py`, on every changed source file; `git
diff --check` also passed.
**Added but not executed here:** `ColumnarBatchOutIteratorExceptionSuite`
(iterator-hook/schema/fallback coverage), native cast-attribution and
legacy/TRY cases in `SparkFunctionTest`, and Velox integration cases covering
the eight shapes under both LEGACY and ANSI assignment policies plus legacy/TRY
results in `VeloxCastSuite`.
**Remaining verification:** the native additions, full native/JVM reactor
builds, and the original **16 Delta streaming tests have not been run**. A
bounded native configuration attempt reached GCC/CMake/Ninja but stopped
because the pinned Velox checkout/build dependencies are absent
(`velox_ep/scripts/setup-helper-functions.sh` is missing). This is not a claim
that the native build or all 16 streaming cases are green.
On this Windows checkout, the unscoped Scala formatter encounters a
pre-existing symlink checked out as path text in the Delta sources; scoped
formatting of this patch passes. The documented `dev/check.py` entrypoint also
expects the absent `dev/license-header.py`; the repository's actual CI header
checker was run directly instead. Formatter-only line-ending changes and
temporary build artifacts were cleaned up before committing.
## Was this patch authored or co-authored using generative AI tooling?
<!--
If generative AI tooling has been used in the process of authoring this
patch, please include the
phrase: 'Generated-by: ' followed by the name of the tool and its version.
If no, write 'No'.
Please refer to the [ASF Generative Tooling
Guidance](https://www.apache.org/legal/generative-tooling.html) for details.
-->
Generated-by: GitHub Copilot 1.0.87-0 (GPT-6 Astra)
--
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]
