Help needed in the query.

Question

I have a table say T1 with a column of unique sequential numbers that begin at 1. When one inserts a new row, one muse use a sequence number that is not currently in the column. that is Fill the gaps. If there are no gaps, then and only then one can use the next highest integer in the sequence.
create table T1
( seq_nbr INTEGER NOT NULL PRIMARY KEY
CHECK (seq > 0);
junk CHAR(5) NOT NULL);
INSERT INTO T1
VALUES (1,'Tom'), (2,'Dick'), (4,'Harry'), (5,'Moe');
Here is the INSERT statement,i came across, where i am not able to understand how logic is working?
INSERT INTO T1 (seq_nbr,junk)
VALUES (CASE WHEN EXISTS -- no gaps
( SELECT 'no gaps'
FROM T1
HAVING COUNT(*) = MAX(seq_nbr))
THEN ( SELECT MAX(seq_nbr) FROM T1 ) + 1
ELSE ( SELECT MIN(seq_nbr) -- gaps
FROM T1
WHERE (seq_nbr -1)
NOT IN (SELECT seq_nbr FROM T1)
AND seq_nbr > 0) - 1, 'Jen');
the thing is i am not able to understand how this query is executing and the logic.. Can someone please explain this to me?

Answer 1

Hi,
this works because the first part of the CASE statement assumes there are no gaps if the number of records in the table = the max(seq_nbr) and inserts the next one up. If there are gaps then it looks for the lowest missing seq_nbr. (without testing this code I wouldn't assume it works)
Overall though this is not good practice because with high volume systems it will most likely lead to all sorts of concurrency problems. Instead you should just use a straight oracle sequence. Every time I've seen someone try to specify the requirement that a PK must be contiguous set of numbers with no gaps I've found it to be based on an individuals perverse sense of completeness or a left over from an old manual system. Usually the sequence numbers where generated from a ledger and the only way they could easily count throughput was by subtracting yesterdays closing sequence from todays. Once you point out that the database can count these things in fractions of seconds, the requirement disappears.
Andre