Variable in subquery not using index in timestamp field

Question

I'm trying to optimize a query but when I use a variable in a subquery, the index is not being used.

set @dh = '2018-01-17 23:59:59'
...
inner join cons c1 on c1.idcons = xx.maxcons
left join conslog clog on clog.idconslog = (select max(clt.idconslog)
                                              from conslog clt
                                              where clt.idcons = c1.idcons
                                              and clt.date_create <= @dh)
...

I get with explain

+----+-------------+-------+------+---------------+-----+---------+-----+----------+----------------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref |   rows   |                       Extra                        |
+----+-------------+-------+------+---------------+-----+---------+-----+----------+----------------------------------------------------+
|  1 | PRIMARY     | clog  | ALL  |               |     |         |     | 40978775 | Using where; Using join buffer (Block Nested Loop) |
+----+-------------+-------+------+---------------+-----+---------+-----+----------+----------------------------------------------------+

If, instead of using a varible, a run the query replacing it with the string, like:

...
inner join cons c1 on c1.idcons = xx.maxcons
left join conslog clog on clog.idconslog = (select max(clt.idconslog)
                                              from conslog clt
                                              where clt.idcons = c1.idcons
                                              and clt.date_create <= '2018-01-17 23:59:59')
...

explain gives me:

+----+-------------+-------+--------+---------------+---------+---------+------+------+-------------+
| id | select_type | table |  type  | possible_keys |   key   | key_len | ref  | rows |    Extra    |
+----+-------------+-------+--------+---------------+---------+---------+------+------+-------------+
|  1 | PRIMARY     | clog  | eq_ref | PRIMARY       | PRIMARY |       4 | func |    1 | Using where |
+----+-------------+-------+--------+---------------+---------+---------+------+------+-------------+

I've checked other answers here on SO, tried cast ing the variable, convert_tz to UTC, creating it using timestamp(date, time), date_format ... I'm running out of ideas.

date_create is of type:

date_create          TIMESTAMP DEFAULT CURRENT_TIMESTAMP         NOT NULL,

Why does this happen? Why does it have to check so many rows since I'm using idcons which is PK?

Thank you for your help

Answer 1

How about this instead, to simplify the only get the records you need. I think the clog.idcons=c1.idcons might help.

I think it might also help to change the nested query to use clog since that's the one associated with the nested =.

left join conslog clog on clog.idcons=c1.idcons and clog.idconsumolog = (select max(clt.idconslog)
                                              from conslog clt
                                              where clog.idcons = clt.idcons
                                              and clt.date_create <= @dh)