Re: sequential scan unduly favored over text search gin index

On Mon, Jun 20, 2011 at 8:38 AM, Sushant Sinha <sushant354(at)gmail(dot)com> wrote:
>
> postgres version 9.0.2
> statistics on docvector is set to 10000 and as you can see the row
> estimates are fine.
>
> lawdb=# \d docmeta1
>       Table "public.docmeta1"
>   Column    |   Type    | Modifiers
> -------------+-----------+-----------
>  tid         | integer   | not null
>  docweight   | integer   |
>  doctype     | integer   |
>  publishdate | date      |
>  covertids   | integer[] |
>  titlevector | tsvector  |
>  docvector   | tsvector  |
> Indexes:
>    "docmeta1_pkey" PRIMARY KEY, btree (tid)
>    "docmeta1_date_idx" btree (publishdate)
>    "docmeta1_docvector_idx" gin (docvector)
>    "docmeta1_title_idx" gin (titlevector)
>
> lawdb=# SELECT relpages, reltuples FROM pg_class WHERE relname
> ='docmeta1';
> relpages | reltuples
> ----------+-----------
>    18951 |    329940

What the are sizes of associated toast tables for the tsvector columns?

>
> lawdb=# explain analyze select * from docmeta1 where docvector @@
> plainto_tsquery('english', 'free');

It would be nice to see the results of explain (analyze, buffers).

Cheers,

Jeff