Alvaro Herrera wrote:
> While we're at this, let's consider Heikki's patch for the streaming
> indexscan API stuff. That patch was supposed to come from the bitmap
> index patch, but it was also supposed to help the Grouped Index Tuples
> (GIT) patch.
>
> In fact, as far as I understood the discussion, GIT would be helped by
> the streaming indexscan API patch, because it would help de-uglify
> certain parts of that patch. It is my impression that we would not want
> to commit the ugly GIT patch; so it would mean that either we commit the
> streaming indexscan patch first, and a beautiful GIT patch shortly later,
> or we don't commit any of them.
Right, for GIT the relevant part of the patch is the support for
returning candidate matches.
However, that's just one source of ugliness in the GIT patch. The
amgettuple interface needs to be changed as well, here's the proposal
for that:
http://archives.postgresql.org/pgsql-hackers/2007-03/msg01079.php
The internals aren't that beautiful either, but I wanted to start from
the API.
> So, the question is: do we want the GIT patch in 8.3? If we do, then we
> need the streaming indexscan API patch. And thus we would need an
> updated (beautiful) GIT patch.
>
> Is there a resolution on this? Alexey Kluykin and I would be interested
> in helping review this set of patches, if it is decided that they are
> wanted for 8.3.
Thanks!
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com