> Currently, one issue you're going to face is that brin doesn't rescan a
range to
> find the tighest possible summary tuple.
That's going to be an issue I think, thanks for mentioning it. We'd need
some sort of mechanism for achieving this without a complete REINDEX, even
if it only reset the min
Jack Douglas wrote:
> > If the values are perfectly clustered, the index is optimal because you
> scan the minimal set of pages.
>
> That's the bit I'm particularly interested in, as my plan would be to keep
> the pages well clustered: http://dba.stackexchange.com/a/66293/1396
>
> Do you see any
> If the values are perfectly clustered, the index is optimal because you
scan the minimal set of pages.
That's the bit I'm particularly interested in, as my plan would be to keep
the pages well clustered: http://dba.stackexchange.com/a/66293/1396
Do you see any blocker preventing BRIN being used
Jack Douglas wrote:
> > in 9.4, GIN indexes are pretty close to this already
>
> Do I understand correctly that BRIN indexes will be even closer to this?
>
Yeah, in a way. You could say they are closer from the opposite end.
There is one index tuple in a BRIN index for each page range (contiguo
sql-general@postgresql.org
Subject: Re: [GENERAL] new index type with clustering in mind.
Martijn van Oosterhout writes:
> On Sat, May 24, 2014 at 05:58:37PM +0100, Jack Douglas wrote:
>> Would the following be practical to implement:
>> A btree-like index type that points to *
> > > To reduce complexity (eg MVCC/snapshot related issues), index entries
> > > would be added when a row is inserted, but they would not be removed
> > > when the row is updated/deleted (or when an insert is rolled back):
> > It's an interesting idea, but, how can you *ever* delete index ent
> The discussions at PGCon pointed out that with the posting-list
compression logic added in 9.4, GIN indexes are pretty close to this
already. Multiple items on the same heap page will typically only take one
byte of index space per item; but there is an identifiable entry, so you
don't get into
Martijn van Oosterhout writes:
> On Sat, May 24, 2014 at 05:58:37PM +0100, Jack Douglas wrote:
>> Would the following be practical to implement:
>> A btree-like index type that points to *pages* rather than individual rows.
> It's an interesting idea, but, how can you *ever* delete index entries?
On Sat, May 24, 2014 at 05:58:37PM +0100, Jack Douglas wrote:
> Would the following be practical to implement:
> A btree-like index type that points to *pages* rather than individual rows.
> Ie if there are many rows in a page with the same data (in the indexed
> columns), only one index entry wil