Question

With this query

SELECT
o1.created,
SEC_TO_TIME(TIMESTAMPDIFF(second, o1.created, (SELECT MAX(created) FROM orders o2 WHERE o2.created < o1.created and o2.created >= '2012-06-01'))) tsd
FROM
orders o1
WHERE o1.created >= '2012-06-01'
ORDER BY tsd DESC LIMIT 100

I want to find out, if there were any significant times in which no orders were placed and how long this timespan was.

Here’s the explain:

*************************** 1. row ***************************
           id: 1
  select_type: PRIMARY
        table: o1
         type: range
possible_keys: created
          key: created
      key_len: 4
          ref: NULL
         rows: 2649278
        Extra: Using where; Using index; Using filesort
*************************** 2. row ***************************
           id: 2
  select_type: DEPENDENT SUBQUERY
        table: o2
         type: range
possible_keys: created
          key: created
      key_len: 4
          ref: NULL
         rows: 2649278
        Extra: Using where; Using index

created is of type timestamp.

The query runs for over an hour now. Changing anything in the schema isn’t an option. This is a one-time query, but I’m just curious if performance can be boosted.

Answer 1

Try to use this query. Is it faster?

SQLFiddle demo

select created,Previous_created,
       IF(tsd<>0,SEC_TO_TIME(tsd),0) as diff 
from
(
 select created,
       IF(@prev<>0,TIMESTAMPDIFF(second, @prev,created),0) tsd,
       @prev Previous_created,
       @prev:=created f2 
    from orders,(select @prev:=0) t1 
WHERE created >= '2012-06-01'
order by created
 ) t2 
order by tsd desc limit 100