SQL - Remove Duplicate value between two columns - sql

Question

I'm looking a simple way to remove an unwanted Duplicate value.
The Dupe is part of a reference to another column, and not within the column itself, but the column I want to remove the dupe value from is multi-delimited with other values.
Here is an example table:
ID,Thing
Dog,Cat;Dog;Bird
Snake,Horse;Fish;Snake
Car,Car;Bus;Bike
As you can see Dog,Snake,Car are the values I need to remove from the Thing column.
Output:
ID,Thing
Dog,Cat;Bird
Snake,Horse;Fish
Car,Bus;Bike
Is there a way to match within a multidelimited field and pull out the exact match?
I'm using SQL Server MGMT studio. Thanks.

Answer 1

WITH CTE AS
(
SELECT ID, Thing, ROW_NUMBER() OVER (PARTITION BY Thing) AS rn
)
DELETE
FROM CTE
WHERE rn > 1
I believe this will do it. Test first by running just the CTE part of the query so you can see what rn is.

Answer 2

Your question and sample data is not very clear. I think what you want is to remove anything from the second column that is in the first column, in which case you can try using replace
select Id,
replace(replace(thing,id,''),';;',';')
from table
Storing multi-value elements in a column is never a good idea and is a conflict of interest with the relational data model; it pretty much always causes problems at some point.

Answer 3

What you can do is concatenate a leading and a trailing ; to the value of Thing and then replace the value of ID with an empty string.
Then remove the leading and trailing ;.
If your version of SQL Server is 2017+, you can use the function TRIM():
SELECT Id,
TRIM(';' FROM REPLACE(';' + Thing + ';', ';' + ID + ';', ';')) Thing
from tablename;
For previous versions use SUBSTRING():
SELECT Id,
SUBSTRING(
REPLACE(';' + Thing + ';', ';' + ID + ';', ';'),
2,
LEN(REPLACE(';' + Thing + ';', ';' + ID + ';', ';')) - 2
) Thing
from tablename;
If you want to update the table:
UPDATE tablename
SET Thing = TRIM(';' FROM REPLACE(';' + Thing + ';', ';' + ID + ';', ';'));
or:
UPDATE tablename
SET Thing = SUBSTRING(
REPLACE(';' + Thing + ';', ';' + ID + ';', ';'),
2,
LEN(REPLACE(';' + Thing + ';', ';' + ID + ';', ';')) - 2
);
See the demo.

Answer 4

I don't really understand what "multi-delimited" means with respect to a string. In your context it seems to suggest that you might have different types of delimiters. It definitely does mean that you have a really poor data model. If you want to remove the id from the things column, then my first suggestion is to fix the delimiters.
In SQL Server, you could use:
select t.*,
(select string_agg(s.value, ';')
from string_split(replace(t.things, ',', ';'), ';') s
where s.value <> t.id
) as new_things
from t;
If the delimited have some intrinsic meaning (did I mentioned that you should fix the data model?), then you can use a more brute force approach. Here is one method:
select t.*,
(case when things = id then ''
when things like concat(id, '[,;]%')
then stuff(things, 1, len(id) + 1, '')
when things like concat('%[,;]', id)
then left(things, len(things) - len(id) - 1)
when things like concat('%[,;]', id, '[,;]%')
then stuff(things, patindex(concat('%[,;]', id, '[,;]%'), things), len(id) + 1, '')
else things
end)
from t;
Here is a db<>fiddle.

Answer 5

Your Question is a good one. I used simple case statement to get the answer. It is CHARINDEX that helped to find the location of the value in Id column and then identified the position of the value in id and according to the position, string was replaced by required values.
--Preparing the table
SELECT *
INTO t
FROM (VALUES
('Dog', 'Cat;Dog;Bird'),
('Snake', 'Horse;Fish;Snake'),
('Car', 'Car;Bus;Bike')
) v(id, things)
--Query
SELECT id
,CASE WHEN CHARINDEX(reverse(id), reverse(things), 1) = 1 THEN REPLACE(things,';'+id ,'')
WHEN CHARINDEX(id, things, 1) < LEN(things) AND CHARINDEX(id, things, 1) > 1 THEN REPLACE(things, id +';' ,'')
WHEN CHARINDEX(id, things, 1) = 1 THEN REPLACE(things, id +';' ,'')
ELSE 'End'
END as [things]
FROM t