Re: Incremental sort for access method with ordered scan support (amcanorderbyop)

On Wed, 19 Apr 2023 at 16:53, Miroslav Bendik <miroslav(dot)bendik(at)gmail(dot)com> wrote:
> > 2. You can use list_copy_head(root->query_pathkeys,
> > list_length(orderbyclauses)); instead of:
> >
> > + useful_pathkeys = list_truncate(list_copy(root->query_pathkeys),
> > + list_length(orderbyclauses));
>
> This code will crash if query_pathkeys is NIL. I need either modify
> list_copy_head (v3.1) or add checks before call (v3.2).
>
> I don't know if it's a good idea to modify list_copy_head. It will add
> additional overhead to every call.

That's a bug in list_copy_head(). Since NIL is how we represent empty
Lists, crashing on some valid representation of a List is not how it
should work.

That function is pretty new and was exactly added so we didn't have to
write list_truncate(list_copy(...), n) anymore. That gets pretty
wasteful when the input List is long and we only need a small portion
of it.

I've just pushed a fix to master for this. See [1]. If you base your
patch atop of that you should be able to list list_copy_head() without
any issues.

David

[1] https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit;h=e35ded29566f679e52888a8d34468bb51bc78bed