Question

I have a database table with a UNIQUE key on a date column so that there can never be more than one row per date. If I insert a new row for a day that already exists in the database, I’d like it to push all the subsequent, consecutive dates back by one day until it reaches a ‘free’ day.

Here’s what I thought:

INSERT INTO
 `activity`
 (`name`,`date`)
VALUES
 ('date 7','2009-07-31')
ON DUPLICATE KEY
 UPDATE `date` = DATE_ADD(`date`, INTERVAL '1' DAY)

My thinking is that the ON DUPLICATE KEY UPDATE will bubble up the table and keep adding one day to every date until it reaches a day that doesn’t exist.

E.g., if my table content looks like this:

date 1, 2009-07-30
date 2, 2009-07-31
date 3, 2009-08-01
date 4, 2009-08-02
date 5, 2009-08-04 <- notice this date is two days past the previous one

…and I go to insert ‘date 7′ on 2009-07-31, I’d like it to do this:

date 1, 2009-07-30
date 7, 2009-07-31 <- newly inserted row
date 2, 2009-08-01 <- this and subsequent rows are all incremented
                      by one day until it hits a non-consecutive day
date 3, 2009-08-02
date 4, 2009-08-03
date 5, 2009-08-04 <- notice this date hasn't changed

But on duplicate key update doesn’t work like that, it only updates the clashing row, and if that row then clashes with a different row, it bails out with a duplicate key error.

Is there a smart way to achieve this effect in SQL alone?

Answer 1

Query for the first free date on or after your target date. This involves a left self-join to find dates with no successor in the table.

SELECT DATE_ADD(Min(a.`date`), INTERVAL '1' DAY) AS `free_date`
FROM `activity` a
LEFT JOIN `activity` z
ON z.`date` = DATE_ADD(a.`date`, INTERVAL '1' DAY)
WHERE z.`date` IS NULL 
AND a.`date` >= '2009-07-31'

Run an update to increment each date in the range between your target date and the first free date.

Now there is room for your desired insert.