wildcard search performance with "like"

От: Yantao Shi
Тема: wildcard search performance with "like"
Дата: ,
Msg-id: 43CD4CE5.3020506@larcb.ecs.nasa.gov
(см: обсуждение, исходный текст)
Ответы: Re: wildcard search performance with "like"  (Tom Lane)
Re: wildcard search performance with "like"  (Michael Riess)
Список: pgsql-performance


I have a postges 8.1.1 table with over 29 million rows in it. The colunm
(file_name) that I need to search on has entries like the following:



I have an index on this column. But an index search is performance only
when I give the full file_name for search:

testdbspc=# explain select file_name from catalog where file_name =
Index Scan using catalog_pk_idx on catalog  (cost=0.00..6.01 rows=1
  Index Cond: (file_name =
(2 rows)

What I really need to do most of the time is a multi-wildcard search on
this column, which is now doing a whole table scan without using the
index at all:

testdbspc=# explain select file_name from catalog where file_name like
Seq Scan on catalog  (cost=0.00..429.00 rows=1 width=404)
  Filter: (file_name ~~ 'MOD04_L2.A2005311.%.004.2005312013%.hdf'::text)
(2 rows)

Obviously, the performance of the table scan on such a large table is
not acceptable.

I tried full-text indexing and searching. It did NOT work on this column
because all the letters and numbers are linked together with "." and
considered one big single word by to_tsvector.

Any solutions for this column to use an index search with multiple wild

Thanks a lot,
Yantao Shi

В списке pgsql-performance по дате сообщения:

От: Mark Kirkwood
Сообщение: Re: Suspending SELECTs
От: Hari Warrier
Сообщение: Getting pg to use index on an inherited table (8.1.1)