SQL picking one record over another

Question

This is my Original Data:

Attempt Value   Time
1       suspend 1391685762
2       suspend 1391686288
3       suspend 1391686413
4       suspend 1392648082
5       suspend 1392648230
6       suspend 1393255593
7       suspend 1393255953
8       suspend 1393257567
9       suspend 1393257738
9       75      1393341840
10      suspend 1393359336
10      62      1393361876
11      suspend 1393422454
11      87      1393425173
12      suspend 1394622268
13      suspend 1394622275
14      suspend 1394622304
14      85      1394623834
15      suspend 1394622379
16      suspend 1394622441
17      suspend 1394622543
18      suspend 1394622785
19      suspend 1394622968
20      suspend 1394623375

And this is my Target Query Result:

Attempt Value   Time
1       suspend 1391685762
2       suspend 1391686288
3       suspend 1391686413
4       suspend 1392648082
5       suspend 1392648230
6       suspend 1393255593
7       suspend 1393255953
8       suspend 1393257567
9       75      1393341840
10      62      1393361876
11      87      1393425173
12      suspend 1394622268
13      suspend 1394622275
14      85      1394623834
15      suspend 1394622379
16      suspend 1394622441
17      suspend 1394622543
18      suspend 1394622785
19      suspend 1394622968
20      suspend 1394623375

In the original data, there are two records for Attempts 9, 10, 11, and 14. I need a query that looks at the Value field, if it's a number, then "picks" that record over the one with a value of "suspend".

I tried using a group by query like:

SELECT Attempt, min(Value), max(Time)
FROM t
GROUP BY Attempt

But this is not what I needed because max(Time) does not correspond to min(Value). I need the original Value, Time pair.

Can anyone help me with this query?

Answer 1

You could use a sub query to achieve this result.

SELECT Attempt, 
    min(Value) as Value,
    (SELECT Time FROM t as t1 
     WHERE t1.Value = min(t.Value) 
         AND t.Attempt = t1.Attempt) as Time
FROM t as t
GROUP BY Attempt

Answer 2

This is the solution I'd go for, which would in my opinion be the fastest:

SELECT DISTINCT(`attempt`) as `attempt`,
   IF(MAX(`value`)=0,'suspend',MAX(`value`)) as `value`, `time`
FROM(
   SELECT `attempt`, IF(`value`='suspend',0,`value`) as `value`, `time`
   FROM `t`
) as `a`
GROUP BY `attempt`

Answer 3

Here is an option using joins. The following query will render better performance.

SELECT 
DISTINCT
COALESCE(tin.Attempt, t.Attempt) Attempt, COALESCE(tin.Value, t.value) Value, COALESCE(tin.Time, t.time) Time
FROM t
LEFT OUTER JOIN
(SELECT 
    Attempt, Value, Time
 FROM t
WHERE Value <> 'suspend') tin
ON t.Attempt = tin.Attempt AND t.Value <> tin.Value;