Really need help with re-number rows

Question

I am trying to update the Precedence value for each row based on its order in the ORDER BY clause. What I am doing is I allow users to insert rows above or below other rows. This is done by adding 0.5 or -0.5 to the PRECEDANCE value of the row they are inserting above or blow. Once this is done I need to update all rows with a new hole number PRECIDANCE value getting ready for the next insert.
Any help would be great, even an entirely different way!
UPDATE
TABLE_A TBL_A
SET
TBL_A.PRECEDENCE = (ROWNUM * 10)
WHERE
TBL_A.EMP_ID in
select
EMP_ID,
PRECEDENCE
from(
select
TBL_A.EMP_ID,
TBL_A.PRECEDENCE
from
TABLE_A TBL_A
where
TBL_A.OTHER_ID = :THIS_ID
order by
2,
1     
)

Answer 1

On average, though, Andrew's procedure will update half the rows in the table on every insert (assuming that your inserts occur between two rows at random). If there are a lot of rows in the table or if there are a lot of inserts, this may not scale particularly well. It may also cause problems if you have an optimistic locking mechanism in place to handle multiple users updating the table simultaneously.
I'd still do something like
CREATE OR REPLACE PROCEDURE insert_after( p_prior_precedence IN NUMBER,
                                          p_row              IN <<table name>>%rowtype )
AS
  l_next_precedence NUMBER;
BEGIN
  SELECT MIN(precedence)
    INTO l_next_precedence
    FROM <<table>>
   WHERE precedence > p_prior_precedence;
  p_row.precedence := (p_prior_preceddence + l_next_precedence) / 2;
  INSERT INTO <<table>> VALUES p_row;
END;This way, you don't have to update any rows when you do an insert. You don't get integer precedence values, but this strikes me as a reasonable trade-off. If you want to re-number things either at query time
SELECT <<columns>>
       RANK() OVER (PARTITION BY <<something>> ORDER BY precedence) pretty_precedence
  FROM <table name>>or as a nightly job, that would work.
Justin
Distributed Database Consulting, Inc.
http://www.ddbcinc.com/askDDBC