I'm using MySQL with PHP. This is like my table: (I'm using 3 values, but there are more)
id | 1 | 2 | 3 ---+---+---+---- 1 | 3 |12 |-29 2 | 5 |8 |8 3 | 99|7 |NULL
I need to get the greatest value's column name in a certain row. It should get:
id | maxcol ---+------- 1 | 2 2 | 2 3 | 1
Are there any queries that will do this? I've been trying, but I can't get it to work right.
Are you looking for something like the GREATEST function? For example:
SELECT id, GREATEST(col1, col2, col3) FROM tbl WHERE ... Combine it with a CASE statement to get column names:
SELECT id, CASE GREATEST(COALESCE(`1`, -2147483646), COALESCE(`2`, -2147483646), COALESCE(`3`, -2147483646)) WHEN `1` THEN 1 WHEN `2` THEN 2 WHEN `3` THEN 3 ELSE 0 END AS maxcol FROM tbl WHERE ... It's not pretty. You'd do better to follow Bill Karwin's suggestion and normalize, or simply take care of this in PHP.
function findcol($cmp, $arr, $cols=Null) { if (is_null($cols)) { $cols = array_keys($arr); } $name = array_shift($cols); foreach ($cols as $col) { if (call_user_func($cmp, $arr[$name], $arr[$col])) { $name = $col; } } return $name; } function maxcol($arr, $cols=Null) { return findcol(create_function('$a, $b', 'return $a < $b;'), $arr, $cols); } This is a great example of the way normalization helps make query design easier. In First Normal Form, you would create another table so all the values would be in one column, on separate rows.
Since you have used repeating groups to store your values across three columns, you can find the column with the greatest value this way:
SELECT id, IF(col1>col2 AND col1>col3, 'col1', IF(col2>col3, 'col2', 'col3')) AS column_with_greatest_value FROM mytable; The short answer is that there is no simple means to do this via a query. You would need to transpose your data and then determine the largest value that way. So something like:
Select Id, ColumnName, Value From ( Select '1' As ColumnName, Id, [1] As Value From Table Union All Select '2', Id, [2] From Table Union All Select '3', Id, [3] From Table ) As Z Where Exists( Select 1 From ( Select '1' As ColumnName, Id, [1] As Value From Table Union All Select '2', Id, [2] From Table Union All Select '3', Id, [3] From Table ) As Z2 Where Z2.Id = Z.Id Group By Z2.Id Having Max(Z2.Value) = Z.Value ) Order By Id This solution depends on a fixed set of columns where you basically name the columns in the UNION ALL queries. In addition, if you have two columns with identical values for the same Id, you will get duplicate rows.
This query will return the max value regardless of NULLs
SELECT MAX(value) FROM (SELECT 1 column_no, col1 value FROM anotherunamedtable UNION ALL SELECT 2, col2 FROM anotherunamedtable UNION ALL SELECT 3, col3 FROM anotherunamedtable) t If you really need the column number then
SELECT id, (SELECT column_no FROM (SELECT 1 column_no, col1 value FROM anotherunamedtable WHERE id = t.id UNION ALL SELECT 2, col2 FROM anotherunamedtable WHERE id = t.id UNION ALL SELECT 3, col3 FROM anotherunamedtable WHERE id = t.id) s ORDER BY max_value DESC LIMIT 1)) as column_no FROM anotherunamedtable t But I think that the last query might perform exceptionally horrible. (Queries are untested)
In the php side, you could do something like this:
foreach ($rows as $key => $row) { $bestCol = $best = -99999; foreach ($row as $col => $value) { if ($col == 'id') continue; // skip ID column if ($value > $best) { $bestcol = $col; $best = $value; } } $rows[$key]['best'] = $bestCol; } Or something similar...
Forests and trees, here's a trivial and fastest solution (providing I didn't fumble); the expression simply looks for the largest column in the row
SELECT id, CASE COALESCE(col1, -2147483648) >= COALESCE(col2, -2147483648) WHEN CASE COALESCE(col2, -2147483648) >= COALESCE(col3, -2147483648) WHEN true THEN 1 ELSE CASE COALESCE(col1, -2147483648) >= COALESCE(col3, -2147483648) WHEN true THEN 1 ELSE 3 END END ELSE CASE COALESCE(col2, -2147483648) >= COALESCE(col3, -2147483648) WHEN true 2 ELSE 3 END END FROM table t a version with IF() would maybe be more readable, but the above should perform a bit better To deal with NULLS an INT value with minimum of -2147483648 was assumed, the expression could be rewritten to deal explicitly with nulls but would have to branch into 8 different cases and is left as an exercise for the OP.