[jeanlyn]

What changes were proposed in this pull request?

Scope the fix to exactly the failing condition:

1. InStream.UncompressedStream.seek(long): accept desired == length only when no bytes were read for the stream (bytes == null) — i.e. the read planner assigned a zero-byte range to it — and park at EOF. All other seeks keep the existing behavior, so genuinely out-of-bounds seeks on real data still throw.
2. InStream.CompressedStream.seek(long): same condition. It also resets currentCompressedStart, because the buffered chunk is dropped and the cached header position is no longer valid (otherwise a later seek back to that cached position would silently stay at EOF).
3. InStream.UncompressedStream.toString(): null-guard currentRange (pre-existing NPE for streams without backing bytes, which would otherwise mask the real exception message).

Why are the changes needed?

The row group index of an all-null trailing row group legitimately points at the end of the stream (both the Java and C++ writers record such positions). When predicate pushdown (SARG) selects only that row group, StripePlanner.processStream computes a zero-byte read range for the stream (start == end), so the stream is created with bytes == null. UncompressedStream.seek() then only special-cases desired == 0 && bytes == null, and a seek to desired == length falls through the range loop and throws:

Caused by: java.lang.IllegalArgumentException: Seek in column 2 kind DATA to 8192 is outside of the data
    at org.apache.orc.impl.InStream$UncompressedStream.seek(InStream.java:...)

A full scan (no SARG) reads the same file fine, as does the C++ reader. See ORC-2226 for a minimal, self-contained reproduction (also reproduced on 2.1.2 and 2.3.1).

The condition is deliberately narrowed to bytes == null so that no behavior changes for streams that actually have data: for bytes != null, seeking to the stream end keeps the pre-existing semantics (accepted only at the end of the last disk range, rejected otherwise), and corrupt/truncated files still fail loudly instead of silently reading EOF.

How was this patch tested?

New unit tests:

• TestInStream#testUncompressedSeekToEndOfStream — an UncompressedStream with no backing bytes: seek(length) is accepted, read() then returns -1, available() returns 0, and seeking past the end (length + 1) still throws.
• TestInStream#testCompressedSeekToEndOfStream — same for CompressedStream.
• TestInStream#testCompressedSeekBackFromEndOfStream — read, seek to the end, then seek back to the start: the cached chunk-header state must stay consistent and the re-read must return the correct value.
• TestRecordReaderImpl#testSargOnAllNullTrailingRowGroup — end-to-end regression: writes an uncompressed file whose trailing row group is all-NULL for a string column, then reads it with a SARG that selects only that row group; asserts both the row count and that the values are all NULL (fails before the fix, passes after).
• TestRecordReaderImpl#testSargOnAllNullMiddleRowGroup — boundary guard: an all-null row group in the middle of a stripe keeps working via the following row group's estimated end offset.

Full orc-core suite passes locally: 1532 tests, 0 failures, 0 errors.

Was this patch authored or co-authored using generative AI tooling?

Yes, co-authored using generative AI tooling. The analysis, fix and tests were reviewed and verified by a human before submission.

Generated-by: Kimi Code 2.1.1

Notes for reviewers

• Seeks beyond the stream end (desired > length), and any non-zero seeks other than desired == length on streams with no backing bytes, still throw — genuinely corrupt files are not masked.
• Behavior now matches the C++ reader, which tolerates seek-to-end.
• Cross-module impact: none; no file-format or API changes. Subclasses (EncryptedStream, EncryptedCompressedStream) inherit the fix.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

4 open issues

• This test expects an IOException when seeking past the end, but the stream implementations… · New
• These tests use String.format(...) without specifying a locale, which makes output depend on the… · New
• These tests use String.format(...) without specifying a locale, which makes output depend on the… · New
• These tests use String.format(...) without specifying a locale, which makes output depend on the… · New

What changed in this PR

Fixes ORC-2226 by allowing InStream.seek(length) for streams that have no planned read ranges (zero bytes assigned), which is needed when predicate pushdown selects an all-null trailing row group whose index points at end-of-stream.

Changes:

• Allow seek-to-EOF (desired == length) when bytes == null in both UncompressedStream and CompressedStream (and reset compressed cached state).
• Null-guard UncompressedStream.toString() to avoid masking the real exception on zero-byte streams.
• Add unit + regression tests covering seek-to-EOF behavior and SARG selection of all-null row groups.

File	Description
java/​core/​src/​java/​org/​apache/​orc/​impl/​InStream.java	Accept seek(length) for zero-byte planned streams; reset compressed cached state; guard toString() against currentRange == null.
java/​core/​src/​test/​org/​apache/​orc/​impl/​TestInStream.java	Add focused unit tests for seek-to-EOF on uncompressed/compressed streams and seeking back after EOF.
java/​core/​src/​test/​org/​apache/​orc/​impl/​TestRecordReaderImpl.java	Add end-to-end regression tests for SARG selection of all-null trailing/middle row groups.

_{Review effort: Lite}