On 2020-12-09 07:37, Michael Paquier wrote:
>>> Only thing I could think of is that it maybe could use a (small)
>>> comment in the message on that/why get_constraint_index is moved to
>>> utils/lsyscache from catalog/dependency, as that took me some time to
>>> understand.
>>
>> commit message could reasonably say that maybe, but I don't think we
>> need to memorialize it in a comment. lsyscache.c *is* where one
>> would expect to find a simple catalog-field-fetch function like this.
>> The previous implementation was not that, so it didn't belong there.
>
> Agreed.
Thanks, I committed it with an expanded commit message.
After further inspection, I'm not going to do anything about the nearby
get_index_constraint() at this item. The current implementation can use
an index on pg_depend. A scan of pg_constraint has no index available.