On Dec 10, 2014, at 19:38 , Bruce Momjian wrote:
>
> Are you saying when you use a GIN index on a,b,c fields, you can do
> lookups on them independently, like 'c'? I was not aware that works,
> but it might. I know it doesn't work for traditional btree as the index
> is hierarchical. You can l
On Wed, Dec 10, 2014 at 05:27:16PM -0800, Guyren Howe wrote:
> >> Given the futility of database benchmarking in general, I didn’t
> >> want to go any further with this. What I was interested in was
> >> whether it might be worth switching from BTree to GIST/GIN indexes
> >> with regular sorts of d
On Dec 6, 2014, at 12:38 , Bruce Momjian wrote:
>
> On Wed, Dec 3, 2014 at 01:15:50AM -0800, Guyren Howe wrote:
>> GIN is certainly not the “three times” size suggested in the docs, but
>> perhaps
>> that just hasn’t been updated for the 9.4 improvements. Certainly, there
>> isn’t
>> sufficien
On Wed, Dec 3, 2014 at 01:15:50AM -0800, Guyren Howe wrote:
> GIN is certainly not the “three times” size suggested in the docs, but perhaps
> that just hasn’t been updated for the 9.4 improvements. Certainly, there isn’t
> sufficient difference here to make the BTree advantage compelling in most
Obviously, database benchmarking is basically a silly idea, because every
workload, every hardware configuration, every schema are different, with very
different consequences.
Still, I was left with wondering when one should choose a BTree vs GIST or GIN
(I didn’t even try to look at SP-GIST: a
