Spark: Avoid String and java.util.UUID allocations on UUID read/write paths

[wombatu-kun]

What & why

This implements the existing // TODO: direct conversion from string to byte buffer in SparkValueWriters. The Spark data layer converted UUID values through an intermediate String and a java.util.UUID object on every value, in both directions: write did UUID.fromString(s.toString()) then re-serialized the two longs to 16 bytes; read did UUIDUtil.convert(buf).toString() then wrapped that as UTF8String. The UUID arrives as the ASCII bytes of its canonical string and leaves as 16 raw bytes (and vice versa), so the String/UUID objects are pure per-row allocation overhead.

Changes

• Add UUIDUtil.convertToByteBuffer(byte[] uuidStringBytes, ByteBuffer reuse) — parses the 36 ASCII bytes of a canonical UUID string directly into the 16-byte big-endian form.
• Add UUIDUtil.convertToStringBytes(ByteBuffer uuidBytes, byte[] reuse) — renders 16 bytes back to the 36 ASCII bytes of the canonical string.
• Rewire all UUID read/write sites in Avro/Parquet/ORC Spark*Readers/Spark*Writers for Spark 4.1 to use these helpers. The change is identical for 3.5 and 4.0; a backport PR will follow.
• Add TestUUIDUtil coverage for the new methods.

Correctness

Both helpers pivot on the (mostSigBits, leastSigBits) long pair: the parser reproduces java.util.UUID.fromString (parse [0,8) → <<16 | [9,13) → <<16 | [14,18) for MSB; [19,23) → <<48 | [24,36) for LSB) and then putLong(0, msb); putLong(8, lsb) exactly as the previous convertToByteBuffer(UUID, reuse); the formatter is the inverse and matches UUID.toString(). The output is therefore byte-for-byte identical to the previous code. The write side keeps the reusable thread-local buffer; the read side must allocate a fresh array because UTF8String.fromBytes wraps without copying (a reused buffer would alias across rows).

---

AI Disclosure

• Model: Claude Opus 4.7
• Platform/Tool: Claude Code
• Human Oversight: fully reviewed
• Prompt Summary: Implement the "// TODO: direct conversion from string to byte buffer" in SparkValueWriters, removing the String and java.util.UUID allocations on the UUID read/write paths.

[github-actions]

This pull request has been marked as stale due to 30 days of inactivity. It will be closed in 1 week if no further activity occurs. If you think that’s incorrect or this pull request requires a review, please simply write any comment. If closed, you can revive the PR at any time and @mention a reviewer or discuss it on the dev@iceberg.apache.org list. Thank you for your contributions.

[wombatu-kun]

no stale