Chronos/DBWaitlist

Changelog / Chronos DB

Chronos DB 0.5.2

A merge undone in part can be undone again

Beta4 changesDownloads

Shipped with Chronos Gateway 0.3.2.

A merge undone in part can be undone again. Beside Postgres, MySQL, MariaDB or MongoDB, an undo that left rows changed since could never be finished: the rows it had put back read as "changed since" themselves. They are now told apart, inside the undo's own transaction. serve and refresh name the database they work beside, and the help covers all four. Nothing changes in how merges are written. Built and signed on the maintainer's Mac as 0.5.1 was.

  • Undo knows rows already put back. undo_merges (HTTP undo_merge with ids, UNDO MERGE by id) no longer counts a row that is as it was before its merge as "changed since": it is already, neither undone nor skipped, and never refuses the undo. So a merge undone but for rows changed since can be undone again for those rows once they hold what the merge wrote, with the check still inside the undo's own transaction. The reply has already (a count), and rows with status: "already". Undoing a merge twice finds nothing to do instead of failing.
  • serve and chronos refresh name the database a folder is attached to ("refreshed from MongoDB", "column added in MySQL") instead of always Postgres. chronos usage and chronos attach help cover MySQL, MariaDB, MongoDB, --follow, --refresh-every, --lazy-tenant and refresh --tenant.
  • HTTP 500 and 502 answer with their own reason phrases ("Internal Server Error", "Bad Gateway"), not "Error".
  • Docs: the CLI reference gains attach, refresh, studio, their flags and the environment variables; the security page documents ALTER AGENT ... ROTATE TOKEN (it said there was no rotation) and uses the admin token in its examples; the HTTP API and errors pages cover MySQL, MariaDB and MongoDB, the review op, 502 and 507.
Chronos DB 0.5.2 · Chronos