[bwyltcaxe]

No description provided.

[za-arthur]

We had a discussion about amcheck implementation and IMHO we shouldn't verify orioledb tables/indexes inside heap or btree specific checks of amcheck.

[bwyltcaxe]

We had a discussion about amcheck implementation and IMHO we shouldn't verify orioledb tables/indexes inside heap or btree specific checks of amcheck.

Hi!
As I see, the discussion is about tables verification. I agree that integrating the OrioleDB's table verification into verify_heapam is bad practice, but I don't think so for btree_index_check -- we are using btree and it uses kind of same invariants to check and this is the reason why I think it is the lesser of two evils.

I already tried to come up with the best solution: new function (as verify_orioledb_btree or smth like this). I've tried to look up at any other's extension precedents and I've not found any of them -- unfortunately, it looks like we're the pioneers here.

The main reason why I've decided to integrate directly into btree_index_check (except that we're using btree) is because I don't want to multiply boolean flags at pg_amcheck (which is hardcoded for heap).

I am ready to reimplement it using "flag + new function", but I am not sure this is what we should do

[bwyltcaxe]

I think as we're the pioneers here, then another extension-developers will look at us too. If we will clone boolean flags, then they will too; hence, I think we have to come up with more-or-less common solutions, e.g. self-verification index hook as here. At btree_index_check or at smth like tableam_index_check.

Not sure about tables, so I didn't touch them here.