How to generate unique document number ?

Question

Hi All,
We need to generate Document Number where there should be any gap, so that we cannot use sequence. For example : Invoice Number.
one possible approach like : Select MAX(InvoiceNumber)+1 from invoice, what if that table has millions of rows, is this reliable enough ?
How to avoid two user do the same query at same time that lead to dupplicate number ?
Anybody has done it this way in production? or any other approach ?
Thank you for your info,
xtanto

Answer 1

I strongly second Justin's strong suggestion about re-considering the requirement.
Be aware that for any of the schemes suggested to have a chance of working, all access to the table must be through stored procedures that use a SELECT FOR UPDATE on the counter table. No user in the database can have insert privileges on the document table except the owner. The owner's password should be kept in a locked safe and be unknown to anyone. If for some reason the owner needs access to the database, then remove the password from the safe, do what needs to be done, then get some guy off the street to set a new password, write it down and lock it in the safe again.
If you really need a gap-free number, it needsto be another column, separate from the PK. Use a sequence to generate the pk gaps and all, then create a procedure that runs periodically to update the gap free column. Soemthing like:
SQL> CREATE PROCEDURE fix_gap_free AS
  2     l_v NUMBER;
  3     TYPE id_tbl IS TABLE OF number INDEX BY BINARY_INTEGER;
  4     l_ids  id_tbl;
  5     l_id2s id_tbl;
  6  BEGIN
  7     SELECT MAX(NVL(id2,0)) INTO l_v FROM t;
  8     SELECT id, ROW_NUMBER() OVER (ORDER BY id)
  9     BULK COLLECT INTO l_ids, l_id2s
10     FROM t
11     WHERE id2 IS NULL;
12
13     FORALL i IN 1 .. l_ids.COUNT
14        UPDATE t
15        SET id2 = l_id2s(i) + l_v
16        WHERE id = l_ids(i);
17     COMMIT;
18  END;
19  /
Procedure created.
SQL> SELECT * FROM t;
        ID        ID2
         1
         2
         5
         7So, id2 is going to be the gap-free column.
SQL> exec fix_gap_free;
PL/SQL procedure successfully completed.
SQL> SELECT * FROM t;
        ID        ID2
         1          1
         2          2
         5          3
         7          4So first pass is fine. Now lets add a few rows, so the table now looks like:
SQL> SELECT * FROM t;
        ID        ID2
         1          1
         2          2
         5          3
         7          4
         8
         9
        25and generate more gap free number:
SQL> exec fix_gap_free;
PL/SQL procedure successfully completed.
SQL> SELECT * FROM t;
        ID        ID2
         1          1
         2          2
         5          3
         7          4
         8          5
         9          6
        25          7If there may be deletions in the table, and you need to maintain the gap free nature, then just skip the WHERE id2 IS NULL predicate.
John