Check for CREATE/DROP INDEX in schema deltas

[erikjohnston]

As these should be background updates.

(foot-gun)

Spawning from #18439

[MadLittleMods]

Assuming you've tried this out to ensure the matching actually works 👍

Thanks for making a lint to catch this mistake!

[erikjohnston]

Assuming you've tried this out to ensure the matching actually works 👍

Yup, this run shows it checking a valid delta file, and this one shows it failing due to matching a CREATE UNIQUE index

[MadLittleMods]

Still some strange pre-existing logic in this file but as good as before with a new good feature 👍

[reivilibre]

why would we ban DROP INDEX? Afaik that is instant?

[erikjohnston]

It isn't instant for large tables, and takes out aggressive locks, which why there's a DROP INDEX CONCURRENTLY version

[erikjohnston]

Though TBF it is rare for it to actually cause woe

[reivilibre]

maybe it'd be worth saying that, for new tables, it's fine to create the index right now, but you can write the index definition inline in the CREATE TABLE statement

[MadLittleMods]

Good point! The CREATE INDEX syntax is useful but we would need to make the logic more robust to also make sure we see an associated CREATE TABLE in the same delta.

[reivilibre]

but you can write the index definition inline in the CREATE TABLE statement

lies, it seems you can't do this for indices. :<

[erikjohnston]

D'oh. Good catch. Have updated to allow index creation for tables we create in the same PR.

[MadLittleMods]

Label this magic value

Suggested change

	repo.remote().fetch(refspec="develop")
	repo.remote().fetch(refspec=DEFAULT_BRANCH)

Did this spawn from something you saw? I assume GitHub would give you the default branch anyway if you don't specify one?

[erikjohnston]

This was purely to speed things up, so that we only pull that one branch that we care about rather than all branches