Select Matched Pairs from Two Tables - mysql

Question

I need to select matched pairs from two tables containing similarly structured data. "Matched Pair" here means two rows that reference each other in the 'match' column.
A single-table matched pair example:
TABLE
----
id | matchid
1 | 2
2 | 1
ID 1 and 2 are a matched pair because each has a match entry for the other.
Now the real question: what is the best (fastest) way to select the matched pairs that appear in both tables:
Table ONE (id, matchid)
Table TWO (id, matchid)
Example data:
ONE TWO
---- ----
id | matchid id | matchid
1 | 2 2 | 3
2 | 3 3 | 2
3 | 2
4 | 5
5 | 4
The desired result is a single row with IDs 2 and 3.
RESULT
----
id | id
2 | 3
This is because 2 & 3 are a matched pair in table ONE and in table TWO. 4 & 5 are a matched pair in table ONE but not TWO, so we don't select them. 1 and 2 are not a match pair at all since 2 does not have a matching entry for 1.
I can get the matched pairs from one table with this:
SELECT a.id, b.id
FROM ONE a JOIN ONE b
ON a.id = b.matchid AND a.matchid = b.id
WHERE a.id < b.id
How should I build a query that selects only the matching pairs that appear in both tables?
Should I:
Select the query above for each table and WHERE EXISTS them together?
Select the query above for each table and JOIN them together?
Select the query above then JOIN table TWO twice, once for 'id' and once for 'matchid'?
Select the query above for each table and loop through to compare them back in php?
Somehow filter table TWO down so we only have to look at the IDs in matched pairs in table ONE?
Do something totally different?
(Since this is a question of efficiency, it is worth noting that the matches will be quite sparse, maybe 1/1000 or less, and each table will have 100,000+ rows.)

Answer 1

I think I get your point. You want to filter the records in which the pairs exists on both tables.
SELECT LEAST(a.ID, a.MatchID) ID, GREATEST(a.ID, a.MatchID) MatchID
FROM One a
INNER JOIN Two b
ON a.ID = b.ID AND
a.matchID = b.matchID
GROUP BY LEAST(a.ID, a.MatchID), GREATEST(a.ID, a.MatchID)
HAVING COUNT(*) > 1
SQLFiddle Demo

Answer 2

Try this Query:
select
O.id,
O.matchid
from
ONE O
where
(CAST(O.id as CHAR(50))+'~'+CAST(O.matchid as CHAR(50)))
in (select CAST(T.id as CHAR(50))+'~'+CAST(T.matchid as CHAR(50)) from TWO T)
Edited Query:
select distinct
Least(O.id,O.matchid) ID,
Greatest(O.id,O.matchid) MatchID
from
ONE O
where
(CAST(O.id as CHAR(50))+'~'+CAST(O.matchid as CHAR(50)))
in (select CAST(T.id as CHAR(50))+'~'+CAST(T.matchid as CHAR(50)) from TWO T)
and (CAST(O.matchid as CHAR(50))+'~'+CAST(O.id as CHAR(50)))
in (select CAST(T.id as CHAR(50))+'~'+CAST(T.matchid as CHAR(50)) from TWO T)
SQL Fiddle

Answer 3

Naive version, which checks all the four rows that need to exist:
-- EXPLAIN ANALYZE
WITH both_one AS (
SELECT o.id, o.matchid
FROM one o
WHERE o.id < o.matchid
AND EXISTS ( SELECT * FROM one x WHERE x.id = o.matchid AND x.matchid = o.id)
)
, both_two AS (
SELECT t.id, t.matchid
FROM two t
WHERE t.id < t.matchid
AND EXISTS ( SELECT * FROM two x WHERE x.id = t.matchid AND x.matchid = t.id)
)
SELECT *
FROM both_one oo
WHERE EXISTS (
SELECT *
FROM both_two tt
WHERE tt.id = oo.id AND tt.matchid = oo.matchid
);
This one is simpler :
-- EXPLAIN ANALYZE
WITH pair AS (
SELECT o.id, o.matchid
FROM one o
WHERE EXISTS ( SELECT * FROM two x WHERE x.id = o.id AND x.matchid = o.matchid)
)
SELECT *
FROM pair pp
WHERE EXISTS (
SELECT *
FROM pair xx
WHERE xx.id = pp.matchid AND xx.matchid = pp.id
)
AND pp.id < pp.matchid
;