Tom Lane wrote:
> NikhilS <nikkhils@gmail.com> writes:
>>> Looks like pg_get_indexdef is unwell :-(
>
>> yes, it was unwell in the area where the amcanorder was being processed. The
>> attached patch should fix this.
>
> Hm, there is a definitional issue here. Should pg_get_indexdef print
> this stuff at all when colno is nonzero? The header comment says that
> it is to return the column's variable or expression only. The existing
> code suppresses the opclass in this case, which to me suggests that it
> should suppress DESC/ASC as well. Which is not what Nikhil's patch
> does.
>
> Dave, I think we put in this variant of the function for pgAdmin ---
> what does pgAdmin need?
More is better for us - it saves an ugly query that will get uglier if
we need to figure out ASC/DESC here too :-)
I agree that we should have all or nothing though, so I'd like to see
ASC/DESC and opclass please.
/D