How to replace one digit in a number in SQL? - sql

Question

I'm trying to change group of numbers to another one.
Now:
PRACTICE_ID (FK) GROUP_ID ...
0_4_200_400_0.5 4
0_4_300_300_0.5 4
0_4_400_700_0.5 4
0_5_200_400_0.5 5
0_5_900_400_0.5 5
0_5_650_400_0.5 5
Should be:
PRACTICE_ID (FK) GROUP_ID ...
*0_5_200_400_0.5 4
0_5_300_300_0.5 4
0_5_400_700_0.5 4
*0_5_200_400_0.5 5
0_5_900_400_0.5 5
0_5_650_400_0.5 5
I use the query to reach the target:
UPDATE CUSTOM_PRACTICE
SET PRACTICE_ID = REPLACE(PRACTICE_ID, '0_4', '0_5')
WHERE GROUP_ID = 4
In this case I cannot duplicate values. I was trying to use NOT EXISTS clause but didn't get result
How can I rewrite all values which have not duplicate? Like this:
PRACTICE_ID (FK) GROUP_ID ...
0_4_200_400_0.5 4
0_5_300_300_0.5 4
0_5_400_700_0.5 4
0_5_200_400_0.5 5
0_5_900_400_0.5 5
0_5_650_400_0.5 5

Answer 1

UPDATE CUSTOM_PRACTICE
SET PRACTICE_ID = REPLACE(PRACTICE_ID, '0_4', '0_5')
WHERE GROUP_ID = 4 AND REPLACE(PRACTICE_ID, '0_4', '0_5') NOT IN
(SELECT PRACTICE_ID FROM CUSTOM_PRACTICE)

Answer 2

SQL Server has a very convenient function for this called STUFF(). One way of doing what you want uses NOT EXISTS:
UPDATE CUSTOM_PRACTICE
SET PRACTICE_ID = STUFF(PRACTICE_ID, 3, 1, '5')
WHERE PRACTICE_ID LIKE '0_4%' AND
NOT EXISTS (SELECT 1
FROM CUSTOM_PRACTICE CP2
WHERE CP2.PRACTICE_ID = STUFF(PRACTICE_ID, 3, 1, '5') AND
STUFF(PRACTICE_ID, 3, 1, '5')
);
Note that REPLACE(PRACTICE_ID, '0_4', '0_5') does not really do what yo want, because:
REPLACE('0_4_200_400_0.5', '0_4', '0_5')
returns:
'0_5_200_500_0.5'
not:
'0_4_200_400_0.5'
However, I think I like using an updatable CTE, because the logic only appears in one place:
with toupdate as (
select cp.*
(case when practice_id like '0_4%' then stuff(practice_id, 3, 1, '5')
else practice_id
end) as new_practice_id
from custom_practice cp
)
update t
set practice_id = new_practice_id
from (select t.*, count(*) over (partition by new_practice_id) as cnt
from toupdate
)
where practice_id <> new_practice_id and cnt = 1;