Question

So I have a table that looks like this:

cid  |  tid  |  score
-----+-------+-------
 6   |  383  |   88
 2   |  234  |   91
 3   |  685  |   77
 5   |  543  |   97
etc...

What I am trying to do is find an ordinal field pattern in the tid column, then search for other instances of that pattern in the same column. For example:

I query to display all tids of a single cid, ordered by descending score:

cid  |  tid  |  score
-----+-------+-------
 6   |  909  |  100
 6   |  119  |   99
 6   |  221  |   98
 6   |  765  |   97
etc...

A pattern is established in the tid column: 909, 119, 221, 765. Now I want to find other possible instances of that pattern, where the cid value is something else. (Note: different cids can have the same tids, so cid 6 can have tid 909, and cid 4 can too.)

I want to check for the pattern in a second query. The pattern values can be spaced out, but they must appear in order:

                           QUERY 2:    
cid  |  tid  |  score           cid    tid     score
---------------------         ---------------------
 6   |  909  |  100             5    | 909 < | 100
 6   |  119  |   99             5    | 831   |  97
 6   |  221  |   98             5    | 793   |  96
 6   |  765  |   97             5    | 435   |  96
                                5    | 404   |  95
                                5    | 119 < |  94
                                5    | 221 < |  94
                                5    | 765 < |  94

After the comparison, I would like MySQL to return something like:

cid |  tid pattern  | instances (in order)
----+---------------+----------------------
 5      909               1 (or 0)
 5      119               1 (or 0)
 5      221               1 (or 0)
 5      765               1 (or 0)

Is there a reasonable way to do this in MySQL? Any help appreciated.

Answer 1

Gnerally this kind of operation doesn’t work well across rows – Start by pulling it into a single record per cid:

 INSERT INTO summary (c_id, tids)
 SELECT cid, GROUP_CONCAT(tid ORDER BY score DESC)
 FROM yourtable
 GROUP BY cid;

Then you can easily get a list of all combinations where the patterns match or are a superset:

 SELECT a.cid, a.tids, b.cid, b.tids
 FROM summary a, summary b
 WHERE a.cid<>b.cid
 AND a.cid=6
 AND b.tids LIKE CONCAT('%', a.tids, '%');

You might want to have a look at levenstein functions or convolution (hint: if you’re not analysing genomes then have a look at some code which does).