Chronos/DBWaitlist

Changelog / Chronos DB

Chronos DB 0.5.4

Faster scans, `GROUP BY` and one-key joins

Beta1 changeDownloads

Faster scans, GROUP BY and one-key joins. A filter on an integer column is checked as each row is read, a text column read row after row allocates once, and a join on one whole-number column keeps its keys as numbers. Over 200,000-row tables on one core a filtered count is 25% faster, a GROUP BY 23% and the joins 20%; on two cores GROUP BY is now faster than Postgres 18. Answers are the same; nothing changes in how rows are stored or merges are written. The speed tables of 0.5.3 were measured on a busy machine and are measured again here. Built and signed on the maintainer's Mac as 0.5.3 was.

  • Faster scans and one-key joins. A condition that compares an integer column with a whole number (amount > 500) is checked as each row is read, without the expression being worked out; a text column read row after row for a GROUP BY or a join allocates once, not for each row; a hash join on one whole-number column keeps its keys as numbers. On one core, over 200,000-row tables: a filtered count 13.5 to 10.1 ms, GROUP BY with max of a text 24.0 to 18.4 ms, a join + GROUP BY 95 to 75 ms. Answers are the same. BENCHMARKS.md's 2026-11-16 tables (sections 4 and 16.2) are measured again on an idle machine, with what is still slower than Postgres.
Chronos DB 0.5.4 · Chronos