voonhous commented on PR #18583: URL: https://github.com/apache/hudi/pull/18583#issuecomment-4396505551
Pushed an update addressing the review comments. Applied: - Pinned hoodie.compact.inline.max.delta.commits = '5' on all 4 tables (vector, variant, both blob) - Renamed path to externalPath in outOfLineBlobLiteral - embeddingOf now fails with the missing id in the message instead of a bare .get - Extracted val tablePath in the variant test for consistency with the blob tests Skipped: - The sqlStringLiteral path escaping helper. Paths come from the JUnit tmp dir and BlobTestHelpers, so they will never contain quotes. - Sharing a helper between INLINE and OUT_OF_LINE blob tests. The two diverge in the literal builder, the byte assertion, and the post compaction shape check, so a shared helper would need 3 callbacks for 2 callers. @rahil-c on the cast question: yes, keeping it. A bare SQL literal like 0.1 parses as decimal(1,1) in Spark, but VECTOR(3) is backed by array<float>. The cast avoids relying on implicit decimal to float coercion, which is stricter on older Spark versions. Other tests in this file build vectors from Scala Row(Seq(0.1f, ...)) so they do not hit this path. -- 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]
