fix(spark): report the written file size in the commit message

[jackylee-ch]

Rationale for this change

bytesWritten accumulated FieldVector.getBufferSize() per batch — the uncompressed in-memory Arrow size — and handed that to VortexWriterCommitMessage, whose javadoc documents it as "the number of bytes written". For a compressed file the two differ by the compression ratio.

The real value was already available: VortexWriter.finish() returns a summary carrying the file's physical size, and close() was only calling finish() and discarding the result.

What changes are included in this PR?

Take the size from finish().fileSize() and drop the per-batch accumulation, which also removes work from the write path.

• ./gradlew :vortex-spark_2.13:test --tests '…VortexDataSourceWriteTest' --tests '…VortexDataSourceBasicTest' — 15 pass
• same on :vortex-spark_2.12:test — 15 pass
• spotlessCheck (both variants) and javadoc — clean

No new test: the value only becomes observable through a commit message a driver collects, which the suite has no fixture for.

What APIs are changed? Are there any user-facing changes?

None. VortexWriterCommitMessage.bytesWritten() now reports the file size rather than the in-memory size, which is what its javadoc already promised.

AI assistance

Prepared with agentic AI assistance. I read the writer's finalization path and VortexWriter.finish() to confirm the summary is produced there and that close() discarded it.

[codspeed-hq]

Merging this PR will improve performance by 13.04%

⚡ 1 improved benchmark
✅ 1943 untouched benchmarks
⏩ 85 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	decompress[u64, (1000, 16)]	73.3 µs	64.9 µs	+13.04%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing jackylee-ch:spark-commit-message-file-size (7ce0bfe) with develop (4b37237)}

Footnotes

1. 85 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩