Postgres 不是需要很长时间来执行

Question

I have 2 tables location and distance

This query takes a very long time to execute:

SELECT source.name AS source, source.id AS source_id, dest.name AS destination, dest.id AS destination_id
FROM location AS source, location AS dest
WHERE (source.id, dest.id) NOT IN (
    SELECT source_id, destination_id FROM distance
)

Even LIMIT 100 Takes >30 seconds to return results (Total results are ~15k)

The 2 queries individually run almost instantly:

SELECT source.name AS source, source.id AS source_id, dest.name AS destination, dest.id AS destination_id
FROM location AS source, location AS dest

and

SELECT source_id, destination_id FROM distance

Also, modifying the query with EXCEPT fixes the query runtime:

SELECT source.id AS source_id, dest.id AS destination_id
FROM location AS source, location AS dest
EXCEPT
SELECT source_id, destination_id FROM distance

But I want all the 4 columns, not the just 2.

How can I fix this?

Answer 1

The answer is to use WHERE NOT EXISTS ! (reference)

I just modified my query like:

SELECT source.name AS source, source.id AS source_id, dest.name AS destination, dest.id AS destination_id
FROM location AS source, location AS dest
WHERE NOT EXISTS (
    SELECT 1 FROM distance
    WHERE source_id = source.id AND destination_id = dest.id
)

This runs instantly!

Big thanks RhodiumToad on #postgres IRC!

Turns you should never use NOT IN . Its in Postgres' Don't do this:

Don't use NOT IN , or any combination of NOT and IN such as NOT (x IN (select…)) .

(If you think you wanted NOT IN (select …) then you should rewrite to use NOT EXISTS instead.)

Answer 2

Another option is to make use of an anti left join . Also, you want to use proper, explicit join syntax instead of old-school, implicit joins:

SELECT source.name AS source, source.id AS source_id, dest.name AS destination, dest.id AS destination_id
FROM location AS source
CROSS JOIN location AS dest
LEFT JOIN distance dist
    ON  dist.source_id = source.id 
    AND dist.destination_id = dest.id
WHERE dist.source_id IS NULL

For performance, consider the following indexes:

location(id)
distance(source_id, destination_id)

Finally: unless you do expect routes where the starting and ending point are the same, you can use an INNER JOIN instead of a CROSS JOIN :

SELECT source.name AS source, source.id AS source_id, dest.name AS destination, dest.id AS destination_id
FROM location AS source
INNER JOIN location AS dest
    ON source.id <> dest.id
LEFT JOIN distance dist
    ON  dist.source_id = source.id 
    AND dist.destination_id = dest.id
WHERE dist.source_id IS NULL