I have 3 tables:


I want to get the records gotten from march until now of the users registered in march, counting the unique id of their movements with a distinct and the total of it at the end:

SELECT MONTH(mo.ctime) AS month, COUNT(DISTINCT(unique_id))
FROM users us, movements mo,unique_ids id
WHERE us.time_field > '2014-03-01' AND us.time_field < '2014-03-31 23:59:59'
AND mo.time_field > '2014-03-01'
AND us.room_name LIKE 'name'
AND us.room_name=mo.room_name AND us.user=mo.user AND id.id_log=mo.id_log

The thing is that few month ago in other query I realise that adding DATE() to the time_field into WHERE clause makes faster the query, but adding this into the query shown, gives me different results than the initial one.

What's the difference between time_field > 'x' AND DATE(time_field) > 'x' inside WHERE clause?

  • Could it be a time zone issue? TIMESTAMPs has them, DATEs do not.
    – Vérace
    Jun 18, 2014 at 22:37
  • how could the time zone makes this both queries results to be different?
    – tachomi
    Jun 18, 2014 at 22:53
  • Clients in different timezones? It's just a first thought which is why I commented and didn't answer. Look (here) at FROM_UNIXTIME and compare your fields as a first guess - if you're getting different results, there must be a difference in the data somewhere.
    – Vérace
    Jun 18, 2014 at 22:57
  • mmm ok, if I understood well, all data is located in the same server, so all must have the same timezone
    – tachomi
    Jun 18, 2014 at 23:00
  • Ooops - sorry. It could be that DATE() removes the time and removes/changes/truncates any time field to a default - i.e. midday (or whatever). I can't test as I don't have a running MySQL server at the minute - am installing as we speak. Check it out for yourself.
    – Vérace
    Jun 18, 2014 at 23:04

1 Answer 1


Just addressing this question

What's the difference between time_field > 'x' AND DATE(time_field) > 'x' inside WHERE clause?

The difference is how the Query Optimizer treats them.

  • If the WHERE clause has time_field > 'x', this signals the Query Optimizer to try looking for any index so as to take advantage of it doing a range scan.
  • If the WHERE clause has DATE(time_field) > 'x', this signals the Query Optimizer to throw all indexes under the bus because the DATE function has to be called across the whole table (in the form of a full table scan) or the join result.

Depending on the key distribution of time_field, DATE(time_field) > 'x' triggering a full table scan just happens to be better than a range scan on an index if the values make up a significant percentage of the index. This is even more true using InnoDB because a full table scan passes through the clustered index (where PRIMARY KEY and row data coexist), while a secondary index on time_field would cause a lookup of the secondary index in addition to the clustered index. If such a secondary index had a lopsided key distribution, such an index would be ignored in favor of a full table scan anyway.

This conjuecture is only indicative of your current dataset. Someone else's dataset may have a better (evenly distributed, more balanced) key distribution, resulting in time_field > 'x' working better than DATE(time_field) > 'x'.

To see such differences, run the EXPLAIN on both SELECT queries. The results may be different.

  • Excellent answer as always. Thank you very much.
    – tachomi
    Jun 19, 2014 at 13:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.