String Manipulation in SQL Server (STRING_SPLIT, REVERSE) - sql-server

Question

I want to reverse the sub-strings in NVARCHAR column, that are separated by one character like '-', for example:
Input column cl:
a1-b1-c1-d1
Required output:
d1-c1-b1-a1
I tried REVERSE(cl), the result was 1d-1c-1b-1a!
The best approach I think is using:
STRING_SPLIT(cl,'-')
And then looking for reversing the resulted sub strings and rejoining them, but since we don't know how many delimited sub strings, it is still difficult to handle.
How can we achieve this?
Thank you in advance

Answer 1

REVERSE isn't what you are after here. What you are after is a string splitter that supports (is aware of) ordinal positions; STRING_SPLIT is documented that it explicitly "doesn't care" about the ordinal positions of values in a delimited string.
One function that is aware of ordinal positions is DelimitedSplit8k_LEAD. You can then use that, along with STRING_AGG to recreate your delimited string:
SELECT STRING_AGG(DS.item,'-') WITHIN GROUP (ORDER BY DS.ItemNumber DESC) AS R
FROM (VALUES('a1-b1-c1-d1'))V(S)
CROSS APPLY dbo.DelimitedSplit8K_LEAD(V.S,'-') DS;
Of course, the real solution here is to stop storing delimited data in your database in the first place.

Answer 2

If using an UDF is not an option, you may try a JSON-based approach. You need to transform the input values into a valid JSON array (a1-b1-c1-d1 into ["a1", "b1","c1","d1"]) and parse this array with OPENJSON():
CREATE TABLE Data (ColumnData nvarchar(max))
INSERT INTo Data (ColumnData) VALUES (N'a1-b1-c1-d1')
UPDATE Data
SET ColumnData = (
SELECT STRING_AGG([value], N'-') WITHIN GROUP (ORDER BY CONVERT(int, [key]) DESC)
FROM OPENJSON(CONCAT(N'["', REPLACE(STRING_ESCAPE(ColumnData, 'json'), N'-', '","'), '"]'))
)
Result:
ColumnData
-----------
d1-c1-b1-a1

Answer 3

If you have a modern version of SQL Server (2016 or higher), you can do it as you said: you split the string with string_split, you reverse its order, and aggregate the result with string_agg.
with cte as (
select value,
row_number() over (order by (select null)) as number
from string_split('a1-b1-c1-d1', '-')
)
select string_agg(value, '-') within group (order by number desc)
from cte

Answer 4

This splits the string while preserving the ordinal position without using a custom splitter. To preserve the ordering of words the CHARINDEX function is used as per this.
declare
#string nvarchar(4000)='a1-b1-c1-d1',
#added_delimitter CHAR(1)= '-';
;with ndx_split_cte(split_val, split_ndx) as (
select
sp.[value],
CHARINDEX(#added_delimitter + sp.[value] + #added_delimitter, #added_delimitter + #string + #added_delimitter)
from
string_split(#string, '-') sp)
select string_agg(split_val, '-') within group (order by split_ndx desc) rev_split_str
from ndx_split_cte;
Results
rev_split_str
d1-c1-b1-a1