[core] Support BLOB fields in partial-update tables.

[Stephen0421]

Purpose

This PR adds BLOB support for primary-key tables using the partial-update merge engine.

The main changes include:

• Support scalar blob-descriptor-field with partial updates.
• Support managed blob-field for BLOB, ARRAY<BLOB>, and MAP<K, BLOB>.
• Support scalar blob-view-field resolution and unresolved-reference forwarding.
• Preserve managed BLOB pack dependencies before and after compaction.
• Keep BLOB view prescan consistent with filtering, limits, authorization, and reader configuration.
• Restore global limit behavior across multiple splits.
• Document partial-update, collection, compaction, sidecar, and orphan-pack semantics.

Tests

• SchemaValidationTest
• PartialUpdateMergeFunctionTest
• PrimaryKeyBlobFileWriterTest
• PrimaryKeyPartialUpdateBlobTest
• BlobViewTableReadSupportTest
• PrimaryKeySimpleTableTest#testReadWithLimitThroughTableReadPathMultiSplit
• PrimaryKeySimpleTableTest#testReadWithLimitThroughTableReadPathMultiSplitWithQueryAuthFilter
• PartialUpdateITCase#testBlobDescriptorPartialUpdate
• PartialUpdateITCase#testBlobDescriptorPartialUpdateSequenceGroup
• PartialUpdateITCase#testManagedBlobPartialUpdate
• PartialUpdateITCase#testManagedBlobPartialUpdateSequenceGroup
• PartialUpdateITCase#testBlobViewPartialUpdate
• PartialUpdateITCase#testBlobViewPartialUpdateForwardReference
• PartialUpdateITCase#testBlobViewPartialUpdateSequenceGroup

[JingsongLi]

+1

[JingsongLi]

Why reject managed BLOB and BLOB view fields for partial-update tables?

[Stephen0421]

Why reject managed BLOB and BLOB view fields for partial-update tables?
Actually it can support, already added.

[JingsongLi]

Requesting changes for two correctness issues in the newly enabled managed-BLOB partial-update path. The focused core tests and current CI pass, but neither case below is covered.

[JingsongLi]

Allowing PARTIAL_UPDATE here exposes an incompatible retract path for managed BLOBs. PrimaryKeyBlobExternalizer.externalize replaces every non-null managed BLOB with null for DELETE/UPDATE_BEFORE, while PartialUpdateMergeFunction passes that field to FieldAggregator.retract. With fields.ts.sequence-group=payload and fields.payload.aggregate-function=last_non_null_value, inserting (payload=A, ts=1) and then retracting (payload=A, ts=2) incorrectly keeps A, because a null retract field means ‘keep the accumulator.’ Please either preserve/externalize non-null managed BLOB values on retract (and retain their references), or reject retract-sensitive aggregation for managed BLOB fields. An end-to-end retract regression test would also be needed.

[Stephen0421]

Thanks for catching this. I fixed it at the schema-validation boundary.

For a managed BLOB field protected by a sequence group, retract-sensitive aggregate functions are now rejected because managed BLOB payloads are not retained in retract records. last_value remains supported because it can deterministically clear the field without the original payload, and other aggregate functions can be used with fields.<field>.ignore-retract=true.

The validation still allows safe cases such as ignore-delete=true and managed BLOB fields that are not protected by a sequence group.

I also added end-to-end coverage for newer and older retract sequences, compaction, empty compacted .blobref files, and ignore-retract with scalar, ARRAY, and MAP managed BLOB fields.

[JingsongLi]

This validation only checks the global sequence.field; it misses sequence-group ordering fields encoded in option keys. For example, fields.payload.sequence-group=name is accepted when payload is a managed BLOB in a partial-update table. PartialUpdateMergeFunction.Factory later creates a comparator for payload, and code generation fails with IllegalArgumentException: Illegal type: BLOB. Please parse the key portion of every fields.<key[,key...]>.sequence-group option and reject managed BLOB, ARRAY<BLOB>, and MAP<K, BLOB> fields there, while continuing to allow BLOBs on the protected-value side.

[Stephen0421]

Thanks, fixed. Schema validation now parses the key portion of every partial-update fields.<key[,key...]>.sequence-group option and recursively rejects ordering fields containing BLOB, including ARRAY<BLOB> and MAP<K, BLOB>.

BLOB fields on the protected-value side remain supported. The parser is shared with PartialUpdateMergeFunction.Factory to keep schema validation and runtime behavior consistent.

I also added tests for scalar, ARRAY, MAP, multi-key ordering fields, allowed BLOB protected fields, malformed sequence-group options, and compatibility with non-partial-update merge engines.