Using an order by case in a sub-query

Question

Each schedule has many ramps, and each ramp has an end_date .

I need to list all my ramps, along with an extra field that contains the id of the ramp with the latest end date for that schedule. I also need to be able to use this extra field in my WHERE clause.

This is typically a case for a sub-query using the MAX function, except for one problem: the ramp's end_date field can be null, indicating the ramp is current. Thus, SELECT MAX(end_date) doesn't work, since null values are 'smaller' than non-null values.

So far, this is what I've come up with:

SELECT r1.*,
    (SELECT r2.id
    FROM ramp as r2
    WHERE schedule_id = r.schedule_id
    ORDER BY IF(end_dte is NULL, '9999-99-99', end_dte) DESC
    LIMIT 1) as latestId
FROM ramp as r1

This produces this table, which is exactly what I want:

+-------+-------------+------------+-----------------+--------+------------+------------+----------+
| id    | schedule_id | suppr_flag | comment         | months | start_dte  | end_dte    | latestId |
+-------+-------------+------------+-----------------+--------+------------+------------+----------+
|    16 |           7 | NULL       | NULL            |   NULL | 2008-06-23 | NULL       |       16 |
|    15 |           6 | NULL       | NULL            |   NULL | 2007-05-01 | 2007-12-31 |       15 |
|    13 |           5 | NULL       | 1-15 deals      |   NULL | 2004-08-11 | NULL       |       13 |
|    11 |           4 | NULL       | NULL            |   NULL | 2005-08-11 | NULL       |       11 |
|    12 |           4 | NULL       | NULL            |     12 | 2004-08-11 | 2005-08-10 |       11 |
|    17 |          13 | NULL       | NULL            |      6 | 2009-03-05 | 2009-09-04 |       19 |
|    18 |          13 | NULL       | NULL            |      6 | 2009-09-05 | 2010-03-04 |       19 |
|    19 |          13 | NULL       | NULL            |   NULL | 2010-03-05 | NULL       |       19 |
|    20 |          14 | NULL       | NULL            |     12 | 2001-06-18 | 2008-06-17 |       20 |

except I cannot use latestId in a WHERE clause (it's an unknown column).

Do you have any ideas?

Answer 1

A quick solution would be to repeat the whole query on the ORDER BY clause, because an alias is not seen there, but I don't like much how the query will look:

SELECT r1.*,
    (SELECT r2.id
    FROM ramp as r2
    WHERE schedule_id = r1.schedule_id
    ORDER BY IF(end_dte is NULL, '9999-99-99', end_dte) DESC
    LIMIT 1) as latestId
FROM ramp as r1
ORDER BY
    (SELECT r2.id
    FROM ramp as r2
    WHERE schedule_id = r1.schedule_id
    ORDER BY IF(end_dte is NULL, '9999-99-99', end_dte) DESC
    LIMIT 1);

Or you can SELECT from your original query, and order the result:

SELECT s.*
FROM (
  SELECT r1.*,
      (SELECT r2.id
      FROM ramp as r2
      WHERE schedule_id = r1.schedule_id
      ORDER BY IF(end_dte is NULL, '9999-99-99', end_dte) DESC
      LIMIT 1) as latestId
  FROM ramp as r1
) s
ORDER BY s.latestId

But if I understand your logic correcty, you could use this query to get the maximum end_dte for every schedule_id :

SELECT schedule_id, MAX(COALESCE(end_dte, '9999-12-31')) max_dte
FROM ramp
GROUP BY schedule_id;

Then you can JOIN this query with ramp again to get the ID associated to the maximum end_dte . On the ON clause you will neet to use COALESCE again:

SELECT r1.schedule_id, r2.id as latestId
FROM (
  SELECT schedule_id, MAX(COALESCE(end_dte, '9999-12-31')) max_dte
  FROM ramp
  GROUP BY schedule_id) r1 INNER JOIN
  ramp r2
  ON r1.max_dte = COALESCE(r2.end_dte, '9999-12-31')
     AND r1.schedule_id = r2.schedule_id

And then you can join this again to get the result you need:

SELECT ramp.*, m.latestId
FROM
  ramp INNER JOIN (
    SELECT r1.schedule_id, r2.id as latestId
    FROM (
      SELECT schedule_id, MAX(COALESCE(end_dte, '9999-12-31')) max_dte
      FROM ramp
      GROUP BY schedule_id) r1 INNER JOIN
      ramp r2
      ON r1.max_dte = COALESCE(r2.end_dte, '9999-12-31')
         AND r1.schedule_id = r2.schedule_id
      ) m ON ramp.schedule_id = m.schedule_id
ORDER BY
  latestId

Please see fiddle here . Notice that I am using '9999-12-31' and not '9999-99-99', the first one is a valid date, the second not.

Edit

If you also wish to consider the fact that more than one schedule_id shares the same maximum date, and you just need the latest (maximum) ID in this case, you could use a GROUP BY query, and a MAX aggregated function:

SELECT r1.schedule_id, MAX(r2.id) as latestId
FROM (
  SELECT schedule_id, MAX(COALESCE(end_dte, '9999-12-31')) max_dte
  FROM ramp
  GROUP BY schedule_id) r1 INNER JOIN
  ramp r2
  ON r1.max_dte = COALESCE(r2.end_dte, '9999-12-31')
     AND r1.schedule_id = r2.schedule_id
GROUP BY
  r1.schedule_id

and you use this updated version in your main query.

Answer 2

you could use TWO columns in your order... first for non-null indication flag, then by the date. Change what you had of...

ORDER BY IF(end_dte is NULL, '9999-99-99', end_dte) DESC

to

ORDER BY
   IF( end_dte is null, 1, 2 ),
   end_dte DESC

This way, it pushes all "NULL" end dates to the top of the list (via IF() value would be 1 vs 2 for anything else that HAD a date), THEN, by the date descending