ClickHouse 26.7 ships 61 new features, 112 performance optimizations, and 329 bug fixes. Key highlights include a 313× speedup for GROUP BY ... ORDER BY ... LIMIT queries via a fused streaming pipeline with early termination, three JOIN improvements (build-side granule pruning for 6.2× speedup, packed row references for 12–21% faster hash joins, and a new DPsub join-ordering algorithm), four vector search enhancements (Int8 QBit quantization, strided dimension storage, randomized Hadamard rotations, and quantization codecs for two-stage search), position-aware phrase search for 40× faster text queries, the new EXPLAIN ANALYZE command for measuring actual query execution, unified URL access across S3/GCS/Azure/HDFS, DateTime64 extended to years 0–9999, a new groupFormat aggregate function, and experimental drivers for writing UDFs directly in SQL.

35m read timeFrom clickhouse.com
Post cover image
Table of contents
New contributors #Faster GROUP BY ... ORDER BY ... LIMIT #The right side of a JOIN can now prune the left #Smaller JOIN hash tables with packed row references #DPsub: a third JOIN-ordering algorithm #Faster vector search with QBit #QBit: flexible Int8 quantization #QBit: read fewer dimensions with strided storage #Better quantization with randomized Hadamard rotations #Quantization codecs: fast two-stage vector search #Position for phrase matching #EXPLAIN ANALYZE #Remote and RemoteSecure engines #Unification of URL #DateTime64: Years 0 to 9999 #groupFormat #Drivers for executable UDFs #The built-in Web UI grows into a real SQL workspace #Instant, version-matched reference docs #Regular expressions, compiled to native code #ClickHouse gets faster on macOS #
5.8K Impressions