On 2011-02-28 8:22 PM, Tom Lane wrote:
> Marko Tiikkaja<marko.tiikkaja@cs.helsinki.fi> writes:
>> [ latest version of snapshot-taking patch ]
>
> I started to look at this, and find myself fairly confused as to what
> the purpose is anymore. Reviewing the thread, there has been a lot of
> discussion of refactoring the API of pg_parse_and_rewrite and related
> functions exported by postgres.c; but the current patch seems to have
> abandoned that goal (except for removing pg_parse_and_rewrite itself,
> which doesn't seem to me to have a lot of point except as part of a
> more general refactoring). With respect to the issue of changing
> snapshot timing, most of the discussion around that seemed to start
> from assumptions about the behavior of wCTEs that we've now abandoned.
> And there was some discussion about rule behavior too, but it's not
> clear to me whether this patch intends to change that or not. The
> lack of any documentation change doesn't help here.
>
> So: exactly what is the intended behavioral change as of now, and what
> is the argument supporting that change?
The only intended change is what I was wondering in the original post:
snapshot handling between normal execution and EXPLAIN ANALYZE when a
query expands to multiple trees because of rewrite rules. Like I said
earlier, this is just a bugfix now that wCTEs don't need it anymore.
Rcgards,
Marko Tiikkaja