Update sub-index when inserting in SQL - sql

Question

I have a table that simplified looks like this:
group index value
1 1 text 1
1 2 text 2
1 3 text 3
2 1 text 4
2 2 text 5
2 3 text 6
Group is an foreign key that is also used for grouping items in the table.
Index is an internal index for sorting the items within the group.
Text is just a value.
Then if I do an insert or running a stored procedure to do the insert
INSERT INTO Table VALUES (1, 2, 'new text')
I would like to update the index for the group 1 items so the table looks like this:
group index value
1 1 text 1
1 2 new text (inserted)
1 3 text 2 (index updated)
1 4 text 3 (index updated)
2 1 text 4
2 2 text 5
2 3 text 6
(Running on MS SQL-Server 2008)

Answer 1

This is a store procedure you can use. Takes 3 parameters, grp, idx, and value:
CREATE PROCEDURE ReIndex
#grp int,
#idx int,
#value nvarchar(30)
AS
IF NOT EXISTS(SELECT 1 FROM [tbl] WHERE grp = #grp and idx =#idx)
INSERT INTO tbl VALUES (#grp, #idx, #value)
else
update tbl
set idx = idx+1
where
grp =#grp and
idx >= #idx
INSERT INTO tbl VALUES (#grp, #idx, #value)
Usage:
exec reindex 1, 2, 'new text'
sqlfiddle: http://sqlfiddle.com/#!3/3323d/11

Answer 2

In this answer I'll assume the first column names are grp and idx, as group and index are reserved words in SQL.
You can use the next available idx value by using a MAX(idx)+1 formula, like this:
INSERT INTO t1(grp, idx, value)
SELECT 1, IFNULL(MAX(idx),0)+1, 'new text'
FROM t1
WHERE grp = 1;
But this way two records inserted at about the same time might get the same idx value. To avoid this you could lock your table while inserting the record. You can do that by wrapping the above statement with table lock/unlock statements:
LOCK TABLES t1;
...
UNLOCK TABLES;
However, this method cannot be used within procedures. There the solution involves creating a transaction, and a SELECT ... FOR UPDATE. Note that you need InnoDB Storage for this:
CREATE PROCEDURE ins_val(IN pgrp int, IN pvalue varchar(200))
BEGIN
DECLARE newidx INT;
START TRANSACTION;
SELECT IFNULL(MAX(idx),0)+1
INTO newidx
FROM t1
WHERE grp = pgrp;
FOR UPDATE;
INSERT INTO t1(grp, idx, value)
VALUES (pgrp, newidx, pvalue);
COMMIT;
END;
Note that if you later delete records, these will leave gaps in the idx numbering that will not be filled again by the above method. On the other hand, if the records with the highest idx value is deleted, the next inserted record will re-use that idx value.