PHP mysql Distinct, only load 1 set of ids - php

Question

I'm trying to get data where order doesn't matter with unique ids. So simply my query would be
SELECT DISTINCT id1, id2 FROM messages ORDER BY date
If i have a database with the following data:
id1 | id2 | date
5 | 6 | 1/2/2011
6 | 5 | 1/1/2011
I would only need to load the column with the newest date because the ids are the same 2 people. Really i have to load ids where one of the ids is yours so my real query right now is
SELECT DISTINCT userid_1, userid_2
FROM messages
WHERE userid_2=$dbid
OR userid_1=$dbid
ORDER BY date
and i get a result as [6 5] [5 9] [9 5] [5 15] [5 6] [5 17]
Results 2 and 3 are the same and 1 and 5 are the same. Really only 4 results should be queried. Thanks!

Answer 1

One option is:
SELECT DISTINCT
if(userid_1 >= userid_2,userid_1, userid_2) AS 'id1',
if(userid_1 >= userid_2,userid_2, userid_1) AS 'id2'
FROM messages
WHERE userid_2=$dbid OR userid_1=$dbid ORDER BY date
This query shows two fields for each record.
The first field will be userid1 if it is bigger or equal to userid2, else will show userid2
The second field has the opposite logic
This makes sure that two similar sets of results will alwas be ordered the same way, so the DISTINCT will regard it as same.

Answer 2

If your goal is to get distinct pairings, you could do the following to get the smaller of the pair always in the left column and the larger into the right, thus ensuring that distinct works:
SELECT DISTINCT
(id1 - ( (id1 - id2) + abs(id1 - id2) ) * .5) as first,
(id2 + ( (id1 - id2) + abs(id1 - id2) ) * .5) as second
FROM messages ORDER BY date;
Since the pairs are always arranged, no need for any pairing functions.
Example
+------+------+--------+------------+
| id1 | id2 | val | pair_date |
+------+------+--------+------------+
| 4 | 5 | test 1 | 2010-12-25 |
| 5 | 4 | test 2 | 2011-10-31 |
| 17 | 50 | test 3 | 2011-07-04 |
| 50 | 17 | test 4 | 2001-01-01 |
+------+------+--------+------------+
If I run this query:
SELECT DISTINCT
(id1 - ( (id1 - id2) + abs(id1 - id2) ) * .5) AS first,
(id2 + ( (id1 - id2) + abs(id1 - id2) ) * .5) AS second
FROM pairs ORDER BY pair_date;
I get:
+-------+--------+
| first | second |
+-------+--------+
| 4.0 | 5.0 |
| 17.0 | 50.0 |
+-------+--------+
2 rows in set (0.00 sec)
Obviously using a boolean has the same effect, is easier to read, and probably is faster, but the above algorithm is handy if you just need to swap two numbers so that one side is always the lesser.
I'm not sure why you would have gotten an error. I'm running my tests directly from the MySQL command line....