Question

I have a table in the following format:

row_key  extID  tag   val
-------  -----  ---   ---
1         1     A     a
2         1     A     b
3         1     B     c
4         2     A     d
5         2     C     e

Now I want to have all extID’s where there are several pairs of (tag, val) with specific values, for example:

(tag, val) = (A,a) AND (tag, val) = (B,c)

or,

(tag, val) = (C,e)

The number of constrains can change.

I can think of several ways to do this:

1. Perform a self-join for each constraint
2. Do the searching (iteratively) in the caller program (multiple SQL queries)
3. (Maybe?) write a SQL function to do this
4. Nested SELECT clauses (passing to the outer level the “extID” and using WHERE extID IN (SELECT extID FROM ...)
5. The only true solution that I just can’t find.

Which one would be the preferred (fastest and most elegant) way to do this? (Except, of course, “Surely, 5. is the correct answer.”)

I think a multiple SELF-join is quite elegant. However, I do not know if it is fast and comparatively memory-efficient.

Further, I would like to use a way that works with MySQL, PostgreSQL and SQLite without adaptation – That’s why I can’t use PIVOT afaiu.

Answer 1

SELECT  extID
FROM    tableName
WHERE   (tag = 'A' AND val = 'a') OR
        (tag = 'B' AND val = 'c')
GROUP   BY extID
HAVING  COUNT(*) = 2

• SQLFiddle Demo
• SQL of Relational Division

UPDATE 1

since you haven’t mentioned that there can be duplicate combination of tag and val, DISTINCT keyword is needed.

SELECT  extID
FROM    tableName
WHERE   (tag = 'A' AND val = 'a') OR
        (tag = 'B' AND val = 'c')
GROUP   BY extID
HAVING COUNT(DISTINCT tag, val) = 2

• SQLFiddle Demo