Question

+----------+--------------+-------------------------+
| ticketid | ticketpostid |           date          |
+----------+--------------+-------------------------+
|  1387935 |      3147808 | 2012-09-17 13:33:01     |
|  1387935 |      3147812 | 2012-09-17 13:33:41     |
|  1387938 |      3147818 | 2012-09-17 13:35:01     |
|  1387938 |      3148068 | 2012-09-17 13:37:01     |
|  1387938 |      3148323 | 2012-09-17 14:47:01     |
|  1387939 |      3147820 | 2012-09-17 13:36:01     |
|  1387939 |      3147834 | 2012-09-17 13:36:25     |
|  1387939 |      3147851 | 2012-09-17 13:41:01     |
|  1387939 |      3147968 | 2012-09-17 13:59:06     |
|  1387939 |      3147996 | 2012-09-17 14:03:01     |

This is a result of a query that I wrote. There are two and more than two rows with same ticketid. I need to find the time difference between first two date in each ticketid

Ex.

+----------+--------------+-------------------------+
| ticketid | ticketpostid |           date          |
+----------+--------------+-------------------------+
|  1387935 |      3147808 | 2012-09-17 13:33:01     |
|  1387935 |      3147812 | 2012-09-17 13:33:41     |
|  1387938 |      3147818 | 2012-09-17 13:35:01     |
|  1387938 |      3148068 | 2012-09-17 13:37:01     |
|  1387939 |      3147820 | 2012-09-17 13:36:01     |
|  1387939 |      3147834 | 2012-09-17 13:36:25     |

As a result;

+----------+--------------+
| ticketid |time diff(sec)|
+----------+--------------+
|  1387935 |      40      |
|  1387938 |      120     |
|  1387939 |      24      |

Can you tell me how I can do this?

Thanks.

Answer 1

For PostgreSQL, I think you want the lag window function to compare the rows; it’ll be much more efficient than a self-join and filter. This won’t work with MySQL, as it still doesn’t seem to support the standard SQL:2003 window functions; see below.

To find only the two lowest you can use the dense_rank window function over the ticketid, then filter the results to return only rows where dense_rank() = 2, ie row with the second-from-lowest timestamp, where lag() will produce the row with the lowest timestamp.

See this SQLFiddle which shows sample DDL and output.

SELECT ticketid, extract(epoch from tdiff) FROM (
  SELECT
      ticketid,
      ticketdate - lag(ticketdate) OVER (PARTITION BY ticketid ORDER BY ticketdate) AS tdiff,
      dense_rank() OVER (PARTITION BY ticketid ORDER BY ticketdate) AS rank
  FROM Table1
  ORDER BY ticketid) x
WHERE rank = 2;

I’ve used ticketdate as the name for the date column because date is a terrible name for a column (it’s a data type name) and should never be used; it has to be double quoted in many situations to work.

The portable approach is probably the self-join others have posted. The window function approach above probably works on Oracle too, but doesn’t seem to in MySQL. As far as I can find out it doesn’t support the SQL:2003 window functions.

The schema definition will work with MySQL if you SET sql_mode = 'ANSI' and use timestamp instead of timestamp with time zone. It seems the window functions won’t; MySQL chokes on the OVER clause. See this SQLFiddle.