Skip to content

v4.3.1

Latest

Choose a tag to compare

@mtopolnik mtopolnik released this 29 Oct 10:05
1d8b4c4

Bug fixes for the QWP query client and the QuestDB facade. ILP is unchanged.

Fixes

  • RENAME TABLE, TRUNCATE, SET and similar statements no longer fail. QuestDB 9.4.0 through 10.0.1 report their row count as -1, and the client returned an error after the statement had already run. It now reports 0 rows affected.
  • A statement no longer runs twice after an unreadable server response. The client no longer resubmits a statement after a final response it cannot decode (row count, error, or end of results), a message of unknown type, or a message over its frame size limit.
  • Clearer errors when a statement goes through the wrong method. Query on a non-SELECT says the statement already ran; Exec on a statement that returns rows says to use Query.
  • qdb.Connect accepts query_close_timeout_ms. A connect string containing this key was rejected.

Behavior change

After one of those unreadable responses, a SELECT returns the error instead of failing over. Pooled query clients are evicted automatically; a standalone QwpQueryClient must be closed and rebuilt (#76). A result batch the client cannot decode still fails over, as a dropped connection does.

Known issues

  • #75: with failover on (the default), Query can resubmit COPY ... FROM after a connection drop and start the import twice.
  • #76: a standalone QwpQueryClient stays unusable after a terminal error.

What's Changed

  • fix(qwp): stop RENAME TABLE, TRUNCATE and similar statements from failing or running twice by @mtopolnik in #74

Full Changelog: v4.3.0...v4.3.1