rangareddy commented on PR #19510:
URL: https://github.com/apache/hudi/pull/19510#issuecomment-5305839503
All nine addressed; rebased onto current master first (38 commits, no
conflicts).
**Code:** booleans now derived from column names alone, deleting the
`getReadColumnIDs`/`getReadColumnNames` zip and both its hazards; tests moved
into `TestHoodieParquetInputFormat` as a `@ParameterizedTest` truth table plus
a **reader-level test** through `getRecordReader` — skeleton 3 rows, external
7, nothing projected — which fails `expected: <3> but was: <7>` without the
reorder. 195 tests, checkstyle and rat clean.
**Claims I had wrong, now corrected in the javadoc, body and commit
message:** "Hive 2 never vectorized" (it is the
`hive.vectorized.execution.enabled` default that differs — `iconst_0` vs
`iconst_1`), "applies to both COW and MOR" (MOR only without log files), "zero
coverage" (TestBootstrap/TestOrcBootstrap cover three branches but are
`@Disabled("HUDI-7353")`), and the `INPUT__FILE__NAME` paragraph, which
described a change that never happens.
**Scope:** the body now says this fixes the no-projection shape only. #19643
tracks the external-only and stitch branches that still hand Hive an
out-of-table path, with the vectorization workaround.
Thanks — the reader-level test in particular is the assertion this PR needed.
--
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]
