Question

If I run the following queries each one returns quickly (0.01 sec) and gives me my desired result.

SELECT tagId FROM tag WHERE name='programming'

SELECT COUNT(DISTINCT workcode) FROM worktag WHERE tagId=123 OR tagId=124

(assume the two tagId numbers were the results from the first query)

I would like to combine these queries so I only have to run it once:

SELECT COUNT(DISTINCT workcode) FROM worktag WHERE tagId IN (SELECT tagId FROM tag WHERE name='programming')

However this query completes in about 1 min and 20 sec. I have indexes on worktag.workcode, worktag.tagId, tag.tagId, and tag.name.

If I run DESCRIBE on the queries the first two use the indexes and the second one uses the index for the subquery (on the tag table) but doesn’t use any indexes on the worktag table.

Does anyone know why this might be?

NOTE: the worktag table has over 18 million records in it.

Answer 1

Why don’t you use a join instead of a subquery?

SELECT COUNT(DISTINCT workcode)
FROM worktag
LEFT JOIN tag
  ON worktag.tagId = tag.tagID
WHERE tag.name = 'programming'

P.S.: Seems to be reported as bug.