Re: [BUGFIX] amcanbackward is not checked before building backward index paths

>>>>> "Tom" == Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:

Tom> What amcanbackward is about is whether the index can support
Tom> reversing direction mid-scan, as would be required to support
Tom> FETCH FORWARD followed by FETCH BACKWARD in a cursor. That's
Tom> actually independent of whether the index can implement a defined
Tom> ordering; see for example the hash AM, which sets amcanbackward
Tom> but not amcanorder.

Ugh, so the docs for amutils get this wrong, and if I'd looked at this
more carefully when doing them to begin with I'd have given the
'backwards_scan' property a better name or omitted it entirely.

I'll fix the docs accordingly. I'm referring specifically to this bit:

https://www.postgresql.org/docs/current/static/functions-info.html#FUNCTIONS-INFO-INDEX-PROPS

which I think should say "Can the scan direction be changed in
mid-scan?" in place of the current text (unless anyone has better
wording?)

--
Andrew (irc:RhodiumToad)