[Bug] Full-text search with a row filter returns no rows in full mode when the filter column's scalar index declines the predicate #10380

Search before asking

• I searched in the issues and found nothing similar.

Paimon version
master @ b0af456

Compute Engine
Engine-agnostic (core): Java withFilter, Spark full-text/hybrid search with WHERE.

Minimal reproduce step

1. Data-evolution table, scalar-index.search-mode=full (or detail), full-text index on content, FM index on the filter column (content).
2. Full-text search with withFilter(equal(content, 'paimon zeta')).
3. The result is empty. A plain table scan with the same filter returns the row.

FM always declines =, so DataEvolutionFullTextRead#matchedRows gets no evaluation and drops the covered rows. A full-text index on the filter column behaves the same.

What doesn't meet your expectations?
In full/detail, these rows should be decided from their filter columns (fast unchanged):

• full-text.mdx: full/detail read filter columns of unindexed rows.
• fm.mdx: a declined predicate falls back to the data path.
• DataEvolutionGlobalIndexScanner already treats runtime declines as unindexed ranges ([core] Support composite BTree equality queries for data evolution tables #10339).
• The fast warning recommends full, which currently returns nothing.

Anything else?
N/A

Are you willing to submit a PR?

• I'm willing to submit a PR!