Suppose a table fruits
that looks like this:
------------------------------------------
| id | name | color | calories |
-------------
Here's a way to store values temporarily without using a temp table or a dummy row in your fruit table:
SELECT name, color, calories FROM fruit WHERE id = 2 INTO @name, @color, @calories;
UPDATE fruit AS f1, fruit AS f2
SET
f1.name = f2.name, f2.name = @name,
f1.color = f2.color, f2.color = @color,
f1.calories = f2.calories, f2.calories = @calories
WHERE (f1.id, f2.id) = (2, 5);
Here's another solution that uses a dummy id value:
UPDATE fruit SET id = 0 WHERE id = 5;
UPDATE fruit SET id = 5 WHERE id = 2;
UPDATE fruit SET id = 2 WHERE id = 0;
You could use a join inequality to line up the rows you want to swap:
update fruit a
inner join fruit b on a.id <> b.id
set a.color = b.color,
a.name = b.name,
a.calories = b.calories
where a.id in (2,5) and b.id in (2,5)
http://sqlfiddle.com/#!18/27318a/5
If your operations are based on ID, and you want to swap entire rows, a fancy way of swapping UNIQUE IDs is to start numbering them at 1, and use 0 as a temporary value.
Another way of performing this is using an unsigned column, and using a designated value (ie.: -1) for temporary. I wouldn't really recommend the latter, as we are effectively wasting space with this method. See http://dev.mysql.com/doc/refman/5.0/en/numeric-type-overview.html for more details.
Since ID is unique, it is difficult to just swap the IDs, it's easier to swap the column contents. A query like this might be what you need:
UPDATE
yourtable t1 INNER JOIN yourtable t2
ON (t1.id, t2.id) IN ((1,5),(5,1))
SET
t1.color = t2.color,
t1.name = t2.name,
t1.calories = t2.calories
Please see fiddle here.