Skip to content

Release history

questdb releases

QuestDB is a high performance, open-source, time-series database

All releases

18 shown

Review required
9.4.1 Breaking risk
Breaking upgrade

SQL fuzzer + functions + MVs + Parquet + QWP

No immediate action
9.4.0 Breaking risk

Drop posting index before downgrade

9.3.5 Breaking risk
Breaking changes
  • SAMPLE BY timezone bucket alignment: Sub-day buckets now uniformly spaced in UTC during DST transitions; FROM/TO boundaries interpreted as local time
Notable features
  • Lateral joins with subquery correlation
  • SQL-standard UNNEST for array expansion
  • Statistical window functions: stddev_pop/samp, var_pop/samp, covar_pop/samp, corr
9.3.4 Breaking risk
Breaking changes
  • Parquet Varchar now uses RLE Dictionary encoding instead of Delta Length Byte Array
  • Constant expression folding: Infinity/-Infinity in expressions collapse to NULL
Notable features
  • Dynamic window ranges in WINDOW JOIN
  • Parquet row group pruning with bloom filters
  • Array functions: array_sort, array_reverse, array_elem_min/max/avg/sum
9.3.3 New feature
Security fixes
  • Fixed server crash from HTTP request with overflowing chunk size
Notable features
  • HORIZON JOIN for markout analysis with offset-based aggregations
  • twap() time-weighted average price aggregate
  • WINDOW definition clause for reusable specifications
9.3.2 New feature
Notable features
  • TICK DSL for complex time intervals with timezone and working day support
  • arg_min(), arg_max(), geomean() aggregate functions
  • EMA, VWEMA, percent_rank window functions
9.3.1 New feature
Notable features
  • Arithmetic expressions within window function definitions
  • ksum() window function with Kahan summation
  • Extended tables() metadata with timestamp bounds
9.3.0 New feature
Notable features
  • Views with DECLARE parameters and OVERRIDABLE option
  • PIVOT keyword for row-to-column transformation
  • WINDOW JOIN for time-windowed aggregations
9.2.3 Bug fix
Notable features
  • Faster materialized view updates under contention
  • Faster single-key GROUP BY queries (~2-3×)
  • TTL safety improvements with wall-clock time
9.2.2 Breaking risk
Breaking changes
  • Legacy txn scoreboard v1 removed
  • MANUAL PERIOD views only refresh via manual trigger, not automatically
Notable features
  • Weighted statistics: weighted_avg, weighted_stddev_rel, weighted_stddev_freq
9.2.1 New feature
Notable features
  • Materialized views: REFRESH PERIOD (SAMPLE BY INTERVAL) syntax
  • markout_horizon CROSS JOIN hint for markout analysis
  • New files(), glob() functions for filesystem scanning
9.2.0 New feature
Notable features
  • DECIMAL(precision, scale) with automatic storage optimization (DECIMAL8-DECIMAL256)
  • Support for up to 76 digits of precision
  • Dense ASOF JOIN algorithm for distant row matching
9.1.1 Breaking risk
Security fixes
  • CVE-2023-28642
Notable features
  • Parquet export via COPY TO command
  • Nanosecond precision in Java ILP client
  • Optimized ASOF JOIN for low-frequency symbols
9.1.0 Breaking risk
Breaking changes
  • IN list queries now use hash-based lookups instead of binary search
Notable features
  • TIMESTAMP_NS type for nanosecond precision
  • Symbol column auto-scaling for high-cardinality workloads
  • Continuous profiling with async-profiler integration
9.0.3 Bug fix
Notable features
  • Web Console search across all tabs (Ctrl/Cmd + Shift + F)
9.0.2 New feature
Notable features
  • Parallel ORDER BY + LIMIT execution (up to 10× speedup)
  • Dedicated thread pools for reads, writes, network
  • Array functions: array_stddev, array_min, array_max with negative indexing
9.0.1 New feature

Symbol index performance optimized with 10x–100x speedup for high-concurrency workloads.

9.0.0 Breaking risk
Notable features
  • N-dimensional arrays (DOUBLE[]) with zero-copy slicing and aggregates
  • Materialized view refresh modes: TIMER, MANUAL, PERIOD with calendar awareness
  • Binary line protocol support for DOUBLE[] ingestion

Beta — feedback welcome: [email protected]