如何避免在联合中两次运行昂贵的子查询

Question

I want to union two queries. Both queries use an inner join into a data set, that is very intensive to compute, but the dataset query is the same for both queries. For example:

SELECT veggie_id
FROM potatoes
INNER JOIN ( [...] ) massive_market
    ON massive_market.potato_id=potatoes.potato_id
UNION
SELECT veggie_id
FROM carrots
INNER JOIN ( [...] ) massive_market
    ON massive_market.carrot_id=carrots.carrot_id

Where [...] corresponds to a subquery that takes a second to compute, and returns rows of at least carrot_id and potato_id.

I want to avoid having the query for massive_market [...] twice in my overal query.

Whats the best way to do this?

Answer 1

If that subquery takes more than a second to run, I'd say it's down to an indexing issue as opposed to the query itself (of course, without seeing that query, that is somewhat conjecture, I'd recommend posting that query too). In my experience, 9/10 slow queries issues are down to improper indexing of the database.

Ensure veggie_id, potato_id and carrot_id are indexed

Also, if you're using any joins in the massive_market subquery, ensure the columns you're performing the joins on are indexed too.

Edit

If indexing has been done properly, the only other solution I can think of off the top of my head is:

CREATE TEMPORARY TABLE tmp_veggies (potato_id [datatype], carrot_id [datatype]);

INSERT IGNORE INTO tmp_veggies (potato_id, carrot_id) select potatoes.veggie_id, carrots.veggie_id from [...] massive_market 
    RIGHT OUTER JOIN potatoes on massive_market.potato_id = potatoes.potato_id 
    RIGHT OUTER JOIN carrots on massive_market.carrot_id = carrots.carrot_id;
SELECT carrot_id FROM tmp_veggies
UNION
SELECT potato_id FROM tmp_veggies;

This way, you've reversed the query so it's only running the massive subquery once and the UNION is happening on the temporary table (which'll be dropped automatically but not until the connection is closed , so you may want to drop the table manually).
You can add any additional columns you need into the CREATE TEMPORARY TABLE and SELECT statement

Answer 2

The goal is to pull all repeated query-strings out of the list of query-strings requiring the repeated query-strings. So I kept potatoes and carrots within one unionizing subquery, and placed massive_market afterwards and outside this unification.

This seems obrvious, but my question originated from a much more complex query, and the work needed to pull this strategy off was a bit more involving in my case. For my simple example in my question above, this would resolve in something like:

SELECT veggie_id 
FROM (
  SELECT veggie_id, potato_id, NULL AS carrot_id FROM potatoes
  UNION
  SELECT veggie_id, NULL AS potato_id, carrot_id FROM carrots
) unionized
INNER JOIN ( [...] ) massive_market
  ON massive_market.potato_id=unionized.potato_id 
    OR massive_market.carrot_id=unionized.carrot_id