GH-44344: [Java] fix VectorSchemaRoot.getTransferPair for NullVector

[myegorov]

Rationale for this change

Do not throw UnsupportedOperationException("Tried to get allocator from NullVector") from VectorSchemaRoot.slice() when slicing a VSR containing a NullVector or ZeroVector. Details in #44344

Are these changes tested?

Added unit test that would trigger an UnsupportedOperationException on the legacy path.

• GitHub Issue: [Java] VectorSchemaRoot.slice() throws for NullVector #44344

[myegorov]

I haven't added a test to check this returns null because this seems to me an implementation detail. Please let me know if you think otherwise.

[vibhatha]

I assumed we are going to wait till nullability annotations are properly integrated.

[myegorov]

I assumed we are going to wait till nullability annotations are properly integrated.

I think I understand the value of requiring @Nullable checks, but wondering if the bugfix here should be blocked on the refactor? iiuc the return null in this PR is far from exceptional.

The ability to make zero-copy slices of ValueVector and VectorSchemaRoot is important to our usage of Arrow. Do you think it's possible to fix the bug and file an issue to forbid null by default as a fast follow? I'd be happy to look into this later but might require some rampup.

[vibhatha]

Right. I understand your point. We could create an issue for this and comment in the method.

@lidavidm WDYT?

[lidavidm]

Can we update the base getAllocator docstring to note that some subclasses may not have an allocator? Otherwise I didn't mean that we should block on nullability annotations, just that it would provide an additional layer of checks.

[vibhatha]

Can we update the base getAllocator docstring to note that some subclasses may not have an allocator? Otherwise I didn't mean that we should block on nullability annotations, just that it would provide an additional layer of checks.

Fair enough. Thanks for clarifying @lidavidm

[myegorov]

Any idea of what's causing the protoc-gen-grpc-java errors in https://github.com/apache/arrow/actions/runs/11677293979/job/32514967839?

Error: PROTOC FAILED: /build/java/flight/flight-core/target/protoc-plugins/protoc-gen-grpc-java-1.68.1-linux-x86_64.exe: /lib64/libstdc++.so.6: version `CXXABI_1.3.8' not found (required by /build/java/flight/flight-core/target/protoc-plugins/protoc-gen-grpc-java-1.68.1-linux-x86_64.exe)

Seems like a flaky error? #43264 (comment)

[lidavidm]

Hmm...I merged a gRPC upgrade earlier which had passed CI but maybe it's not quite safe yet. (I recall some newer version of gRPC or its dependencies started requiring newer glibc)

[lidavidm]

Yup, reverting... #44645

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit c3601a9.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 2 possible false positives for unstable benchmarks that are known to sometimes produce them.