voonhous opened a new pull request, #19777:
URL: https://github.com/apache/hudi/pull/19777
… with the row writer
HoodieAvroWriteSupport spliced the forced shredding schema into top-level
fields only and shredded values by top-level index, so a variant nested in a
struct silently declined to the unshredded layout on the AVRO record type while
the row writer shredded it: same table, same config, two on-disk layouts.
Closes #19689.
- VariantSchemaUtils.applyForcedShredding walks records, array elements and
map values the way stripVariantShreddingAt does and force-shreds a variant that
is a record member at any depth, mirroring
HoodieRowParquetWriteSupport.processNestedDataType: a variant that is directly
an array element or map value is not forced on either path (a declared
typed_value still shreds it). Shredded records are named per (enclosing record
type, field) in the hoodie.variant.forced namespace, so a record type reused
under two fields serializes as a name reference instead of tripping Avro's
"Can't redefine" at file open.
- HoodieAvroWriteSupport replaces the index-keyed top-level map with a
Shredder tree built once from the effective schema (the mirror image of
HoodieVariantReconstruction's rebuilder): values are shredded at every position
the effective schema declares typed_value, records match input fields by name
at every level, untouched subtrees are copied by reference. The value walk is
mandatory: parquet-avro serializes positionally, so a spliced nested
typed_value over an untransformed {metadata, value} record fails at write time.
- VariantSchemaUtils.stripVariantShreddingByShape (the TableSchemaResolver
footer fallback) recurses too, so a nested typed_value never leaks into the
resolved table schema, where Hive sync would publish a three-member struct and
both Hive guards would fail open.
- HoodieFileGroupReaderBasedFileFormat.supportBatch applies the top-level
variant policy at any depth: with
spark.sql.parquet.enableNestedColumnVectorizedReader on (off by default)
Spark's ParquetUtils.isBatchReadSupported treats VariantType as atomic and
nested columns as batch-readable, so a nested variant reached the vectorized
reader that #18605 forces off for top-level ones.
Tests: TestHoodieAvroWriteSupportShredding (nested forced reach and the bare
element/value decline, parquet groups under list/element and key_value/value;
red before the fix), TestVariantSchemaUtils (splice reach, identity, naming,
strip-by-shape at depth), TestVariantShreddingMixedLayouts section F (layout
pinned after the AVRO-leg bin-pack merge; record-writer parity leg for both
record types; supportBatch at any depth).
The Spark-native internal read path (compaction, clustering, CDC, MOR merge
over a nested-shredded base) is pinned separately in #19775.
### Describe the issue this Pull Request addresses
<!-- Either describe the issue inline here with motivation behind the
changes
(or) link to an issue by including `Closes #<issue-number>` for
context.
If this PR includes changes to the storage format, public APIs,
or has breaking changes, use `!` (e.g., feat!: ...) -->
### Summary and Changelog
<!-- Short, plain-English summary of what users gain or what changed in
behavior.
Followed by a detailed log of all the changes. Highlight if any code
was copied. -->
### Impact
<!-- Describe any public API or user-facing feature change or any
performance impact. -->
### Risk Level
<!-- Accepted values: none, low, medium or high. Other than `none`, explain
the risk.
If medium or high, explain what verification was done to mitigate the
risks. -->
### Documentation Update
<!-- Describe any necessary documentation update if there is any new
feature, config, or user-facing change. If not, put "none".
- The config description must be updated if new configs are added or the
default value of the configs are changed.
- Any new feature or user-facing change requires updating the Hudi website.
Please follow the
[instruction](https://hudi.apache.org/contribute/developer-setup#website)
to make changes to the website. -->
### Contributor's checklist
- [ ] Read through [contributor's
guide](https://hudi.apache.org/contribute/how-to-contribute)
- [ ] Enough context is provided in the sections above
- [ ] Adequate tests were added if applicable
--
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]
