From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andrew Gierth <andrew@tao11.riddles.org.uk>
cc: pgsql-hackers@lists.postgresql.org, kes-kes@yandex.ru
Subject: Re: mark/restore failures on unsorted merge joins
In-reply-to: <87o8jn50be.fsf@news-spur.riddles.org.uk>
References: <87o8jn50be.fsf@news-spur.riddles.org.uk>
Comments: In-reply-to Andrew Gierth <andrew@tao11.riddles.org.uk>
	message dated "Mon, 23 Nov 2020 19:48:29 +0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <125706.1606161260.1@sss.pgh.pa.us>
Date: Mon, 23 Nov 2020 14:54:20 -0500
Message-ID: <125707.1606161260@sss.pgh.pa.us>
Precedence: bulk

Andrew Gierth <andrew@tao11.riddles.org.uk> writes:
> The problem is that the planner calls ExecSupportsMarkRestore to find
> out whether a Materialize node is needed, and that function looks no
> further than the Path type of T_Index[Only]Path in order to return true,
> even though in this case it's a GiST index which does not support
> mark/restore.

> (Usually this can't be a problem because the merge join would need
> sorted input, thus the index scan would be a btree; but a merge join
> that doesn't actually have any sort keys could take unsorted input from
> any index type.)

Sounds like the right analysis.

> Going forward, this looks like IndexOptInfo needs another am* boolean
> field, but that's probably not appropriate for the back branches; maybe
> as a workaround, ExecSupportsMarkRestore should just check for btree?

Uh, why would you not just look to see if the ammarkpos/amrestrpos fields
are non-null?

			regards, tom lane