A better way than a global temp table to reuse a distinct select?

Question

I get the impression from other threads that global temp tables are frowned upon so I'm wondering if there is a better way to simplify what I need to do. I have some values scattered about a table with a relatively large number of records. I need to distinct them out and delete from 21 other tables where those values also occur. The values have a really low cardinality to the number of rows. Out of 500K+ rows there might be a dozen distinct values.
I thought that rather than 21 cases of:
DELETE FROM x1..21 WHERE value IN (SELECT DISTINCT value FROM Y)
It would be better for performance to populate a global temp table with the distinct first:
INSERT INTO gtt SELECT DISTINCT value FROM Y
DELETE FROM x1..21 WHERE value IN (SELECT value FROM GTT)
People asking questions about GTT's seem to get blasted so is this another case where there's a better way to do this? Should I just have the system bite the bullet on the DISTINCT 21 times? The big table truncates and reloads and needs to do so quickly so I was hoping not to have to index it and meddle with disable/rebuild index but if that's better than a temp table, I'll have to make do.
As far as I understand WITH ... USING can't be used to delete from multiple tables or can it?

Answer 1

Almost, but not quite, as efficient as using a temporary table would be to use a PL/SQL collection and FORALL statements and/or referencing the collection in your subsequent statements). Something like
DECLARE
  TYPE value_nt IS TABLE OF y.value%type;
  l_values value_nt;
BEGIN
  SELECT distinct value
    BULK COLLECT INTO l_values
    FROM y;
  FORALL i IN 1 .. l_values.count
    DELETE FROM x1
     WHERE value = l_values(i);
  FORALL i IN 1 .. l_values.count
    DELETE FROM x2
     WHERE value = l_values(i);
END;or
CREATE TYPE value_nt
  IS TABLE OF varchar2(100); -- Guessing at the type of y.value
DECLARE
  l_values value_nt;
BEGIN
  SELECT distinct value
    BULK COLLECT INTO l_values
    FROM y;
  DELETE FROM x1
   WHERE value = (SELECT /*+ cardinality(v 10) */ column_value from table( l_values ) v );
  DELETE FROM x2
   WHERE value = (SELECT /*+ cardinality(v 10) */ column_value from table( l_values ) v );
END;Justin