Question

select * from ( select * from y where x="t") where z rlike "(.*)query1"
union
select * from ( select * from y where x="t") where z rlike "(.*)query2"

As you can see above, parenthesis contain the same query which typically
returns multi-row results. Is there any way to optimize it by only querying once for the select statement inside the parenthesis and using the result over and over again?

PS: I want to get the results in an ordered manner, so the results of the first unioned query must stay at top.

Answer 1

Three things here:

• First, avoid UNION if you can use UNION ALL. Plain unions involve removal of duplicates and thus take more time.
• When you can factor a query, do it. Here you have union of exact queries except for one criterion, then you can use OR.
• select * from (select * ... is redundancy and can be avoided.

Which gives:

SELECT *
  FROM y
 WHERE x = 't'
   AND ( z RLIKE '(.*)query1'
        OR
         z RLIKE '(.*)query2'
       )

As said in a comment by melpomene (thanks!) you can even factor at the regex level:
WHERE z RLIKE '(.*)query1|(.*)query2'

If there are duplicates in y and you don’t want them, since UNION discarded them, you can introduce DISTINCT but avoid if it is not necessary.

If you want an order, never rely on the DBMS. Use ORDER BY:

ORDER BY CASE WHEN z RLIKE '(.*)query1' THEN 1
              WHEN z RLIKE '(.*)query2' THEN 2
              ELSE 3 -- given the WHERE clause, should never happen
         END