QuestDB

Read it at questdb.com/docs/changelog/

We just added a new docs changelog page to #QuestDB. Release notes tell you what shipped. The docs changelog makes it easier to follow new examples, guides, features, and updates as they land. Link in thread 👇

Read the post at questdb.com/blog/window-jo…

WINDOW JOIN in #QuestDB is not just syntactic sugar. We published a deep dive into how we parallelized and vectorized it, plus benchmarks against Timescale, DuckDB, and ClickHouse. On the tested workload: 5x faster than the previous QuestDB path. Link in thread 👇

A nice thing about designing your own ingestion protocol is choosing exactly what to optimize for. This example from the upcoming QWP protocol in #QuestDB shows multi-table ingestion, schema evolution, Gorilla delta-of-delta timestamp encoding, and symbol dictionary compression.

We are adding more SQL window functions to #QuestDB: • ntile(n) • cume_dist() • nth_value(value, n) Useful for ranking, bucketing, percentiles, and distribution analysis directly in SQL.

For years, we’ve said most #QuestDB workloads do not need indexes. That is still true. But some query shapes benefit from them, especially time filters combined with highly selective predicates. We are now adding covering indexes for those cases. Stay tuned!

We are extending SAMPLE BY FILL in #QuestDB with cross-column PREV() support. Useful for things like OHLC candles, where the next bucket open can inherit the previous close when no trades occurred. Supported queries also move onto the parallel GROUP BY fast path.

Read the post at questdb.com/blog/sparkline…

Last week we mentioned OHLC in #QuestDB. We’ve now written a full blog post on the bigger idea: small charts as text directly in SQL, including sparklines, candlesticks, and market depth charts. Read more in thread 👇

How would you like your query to return OHLC bars directly? Coming soon to #QuestDB: OHLC bars rendered in SQL as a VARCHAR.

With storage policies in #QuestDB Enterprise, the same data can live in native and Parquet at the same time. Fast queries from native, open access via Parquet, with native data evicted automatically as it ages.

Cast your vote at dbta.com/Readers-Choice…

We’re excited to be nominated for the DBTA Readers’ Choice Awards. If #QuestDB has been useful to you, we’d love your support. Voting is open until May 22. Link in thread 👇

Podcast at infoq.com/podcasts/java-…

How do you build a fast database in Java? In this podcast, Jaromir from #QuestDB talks about low-latency design, performance trade-offs, and how AI is starting to change how we develop.

Blog post at questdb.com/blog/code-revi…

Recent work on approx_percentile() in #QuestDB is a good example of how performance improvements actually happen. It started as a solid community contribution, then code review pushed it further, tightening the hot path and improving parallel scaling. Link in the thread🧵

What’s a bloom filter? A compact data structure that tells you if a value is definitely not present, or maybe present. #QuestDB now supports bloom filters on columns when converting to Parquet, so queries can skip row groups and run faster.

With storage policies in #QuestDB Enterprise, you can control the full data lifecycle for tables and materialized views. Convert to Parquet automatically, keep data queryable throughout, and let Parquet continue to accept inserts and updates seamlessly.