CSharpDB v4.0.2
version4.0.2
version4.0.2 improves full-text indexing scalability for large corpora and makes indexed mutations avoid unnecessary full-table scans.
Changed
- Full-text postings are now stored in bounded posting chunks instead of rewriting one growing postings blob for every hot-token insert. This keeps repeated-token indexing scalable for large corpora.
- Existing full-text posting blobs remain readable. Databases with the previous full-text storage layout get the new chunk store on open, and legacy posting blobs migrate to chunked postings on the next write for that term.
- Full-text deletes and updates rewrite only the affected posting chunk while preserving existing search result ordering and transaction rollback behavior.
DELETEandUPDATEstatements can now collect target row IDs through available secondary indexes instead of scanning the full table for indexed predicates.
Validation
- Added focused hot-token before/after benchmark coverage for insert, query, delete, and update paths.
- Added FileSearcher NuGet before/after benchmark automation that validates real package consumption against local baseline and patched CSharpDB packages.