Обсуждение: Partial index

Hi everyone,

i have a function that looks executes the following command in a while loop in which it iterates through tables
(aTable):
FOR rec in EXECUTE 'SELECT count(a.allele_1) as c from aTable a INNER JOIN map_table b on(a.snp_id=upper(b.snp_id))
WHERENOT a.allele_1=a.allele_2 and b.gene=something 

Unfortunately this command leads to 4 minutes of execution for 15 tables of which each has around 3 Million tuples. I
havea partial index on the expression where not allele_1=allele_2 and one on snp_id. 
Here is the explain i get for the above mentioned command:
Aggregate  (cost=229621.08..229621.09 rows=1 width=16)

   ->  Merge Join  (cost=496.29..229361.10 rows=103991 width=16)

         Merge Cond: (("outer".snp_id)::text = "inner"."?column2?")

         ->  Index Scan using idx_snpid_pt1 on snp_allel_chr_11pt1 a  (cost=0.00..212667.07 rows=2875580 width=29)

               Filter: ((allele_1)::text <> (allele_2)::text)

         ->  Sort  (cost=496.29..496.63 rows=138 width=13)

               Sort Key: upper((b.snp_id)::text)

               ->  Bitmap Heap Scan on snps_map b  (cost=2.48..491.38 rows=138 width=13)

                     Recheck Cond: ((gene)::text = 'FN5'::text)

                     ->  Bitmap Index Scan on idx_snps_map_gene  (cost=0.00..2.48 rows=138 width=0)

                           Index Cond: ((gene)::text = 'FN5'::text)
As you can see the partial index is not used. So, does anyone have a suggestion on how to increase the performance of
thequeries. 

Thanks
Chris


Christian Rengstl M.A.
Klinik und Poliklinik für Innere Medizin II
Kardiologie - Forschung
Universitätsklinikum Regensburg
B3 1.388
Franz-Josef-Strauss-Allee 11
93053 Regensburg
Tel.: +49-941-944-7230

Christian Rengstl wrote:
> Hi everyone,
>
> i have a function that looks executes the following command in a while loop in which it iterates through tables
(aTable):
> FOR rec in EXECUTE 'SELECT count(a.allele_1) as c from aTable a INNER JOIN map_table b on(a.snp_id=upper(b.snp_id))
WHERENOT a.allele_1=a.allele_2 and b.gene=something 
>
> Unfortunately this command leads to 4 minutes of execution for 15 tables of which each has around 3 Million tuples. I
havea partial index on the expression where not allele_1=allele_2 and one on snp_id. 
> Here is the explain i get for the above mentioned command:
> Aggregate  (cost=229621.08..229621.09 rows=1 width=16)
>
>    ->  Merge Join  (cost=496.29..229361.10 rows=103991 width=16)
>
>          Merge Cond: (("outer".snp_id)::text = "inner"."?column2?")
>
>          ->  Index Scan using idx_snpid_pt1 on snp_allel_chr_11pt1 a  (cost=0.00..212667.07 rows=2875580 width=29)
>
>                Filter: ((allele_1)::text <> (allele_2)::text)

Well, it's using the index on snp_id instead, and since you're joining I
can see why.

What column(s) do you index with your partial index? If it's snp_id I'd
think it odd that it wasn't used.

--
   Richard Huxton
   Archonet Ltd

"Christian Rengstl" <Christian.Rengstl@klinik.uni-regensburg.de> writes:
> I have a partial index on the expression where not allele_1=allele_2

What is that index's definition *exactly*?  No handwaving please, let's
see the SQL.  Also, what datatypes are these columns?

            regards, tom lane