Reputation: 12564
Need to come up with a way to efficiently execute a query with and array and integer columns in the WHERE
clause, ordered by a timestamp column. Using PostgreSQL 9.2.
The query we need to execute is:
SELECT id
from table
where integer = <int_value>
and <text_value> = any (array_col)
order by timestamp
limit 1;
int_value
is an integer value, and text_value
is a 1 - 3 letter text value.
The table structure is like this:
Column | Type | Modifiers ---------------+-----------------------------+------------------------ id | text | not null timestamp | timestamp without time zone | array_col | text[] | integer | integer |
How should I design indexes / modify the query to make it as efficient as possible?
Thanks so much! Let me know if more information is needed and I'll update ASAP.
Upvotes: 6
Views: 2478
Reputation: 1
Unfortunately GIN index can't be created on integer column so you will have to create two diffrent indexes for those two columns.
That's not entirely true, you can use btree_gin
or -btree_gist
-- feel free to use GIN
CREATE EXTENSION btree_gist;
CREATE INDEX ON table USING gist(id, array_col, timestamp);
VACUUM FULL ANALYZE table;
Now you can run the operation on the index itself
SELECT *
FROM table
WHERE id = ? AND array_col @> ?
ORDER BY timestamp;
Upvotes: 0
Reputation: 16487
PG can use indexes on array but you have to use array operators for that so instead of <text_value> = any (array_col)
use ARRAY[<text_value>]<@array_col
(https://stackoverflow.com/a/4059785/2115135). You can use the command SET enable_seqscan=false;
to force pg to use indexes if it's possible to see if the ones you created are valid. Unfortunately GIN
index can't be created on integer column so you will have to create two diffrent indexes for those two columns.
See the execution plans here: http://sqlfiddle.com/#!12/66a71/2
Upvotes: 2