Repository navigation
fix(postgres): a sequence a column owns but no default uses is created - #293
Merged
Merged
Conversation
A sequence counted as a serial column's own if it fed at most one default, so one feeding none was neither created by its table (no default to write bigserial for) nor by the sequence kind. GitLab's zoekt_tasks fills its id from a trigger and lost its sequence; its p_sent_notifications owns one nothing uses while its default draws on another, and was written bigserial with the wrong sequence. A serial's sequence now feeds exactly one default. A serial whose default was dropped is then standalone on one side and a serial on the other: the sequence kind compares it against the other side's sequence of the same name rather than creating or dropping it, and an OWNED BY change only counts a sequence the other side does not own at all. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.



Two more sequence shapes from GitLab's schema. Both still differed after copying GitLab's schema to an empty database with the previous release.
What was wrong
A sequence counted as a serial column's own if it fed at most one default. A sequence that feeds no default fell through both paths:
Its table didn't create it: with no default, there was nothing to write as
bigserial.The sequence kind didn't create it either, because it treats serials as belonging to their table.
zoekt_tasksfills its id from a trigger (NEW.id := nextval('zoekt_tasks_id_seq')). The column has no default but owns the sequence. The copy had no sequence, so the trigger failed on the first insert. GitLab has ten tables like this.p_sent_notificationsownsp_sent_notifications_id_seq, which nothing uses, while its default draws onsent_notifications_id_seq. The column was writtenbigserial, so the copy's default drew on the owned sequence instead.The fix
OwnedSequences::serialShapednow requires a serial's sequence to feed exactly one default. One that feeds none is standalone: it is created with its options, and itsOWNED BYis set once the table exists.serial_default_droppedcaught that, so two follow-ups keep it working:OWNED BYchange only counts a sequence the other side doesn't own at all, so a serial isn't disowned.Tests
RealWorldShapesPostgresTestgains both shapes, using GitLab's DDL verbatim. Each runs with the pg_dump renderer and with the built-in one. All 4 runs fail on master.Locally:
serial_default_droppedin both directions.PostgreSQL only. MySQL and SQLite have no sequence objects of this kind, and their code paths are untouched.
🤖 Generated with Claude Code