Splitting a column on underscores [duplicate]

Question

This question already has answers here:

Split String Function (2 answers)

T SQL Table Valued Function to Split a Column on commas (5 answers)

How did you split strings before string_split() (3 answers)

Closed 6 months ago.

What SQL syntax would split this column into three new columns (split would be on the underscore)?

There are other columns in the table, but I would like to split these into three (underscore based).

Answer 1

i usually say to people don't save data with any delimiter

CREATE TABLE [test] (JOBNUMBR VARCHAR(15))
GO

INSERT INTO test SELECT '57299_87961_003' UNION SELECT '66515_88567_OOO' UNION SELECT '58OO7_88695_OO1'
GO

SELECT JOBNUMBR
,Reverse(ParseName(Replace(Reverse(JOBNUMBR), '_', '.'), 1)) As [no1]
,Reverse(ParseName(Replace(Reverse(JOBNUMBR), '_', '.'), 2)) As [no2]
,Reverse(ParseName(Replace(Reverse(JOBNUMBR), '_', '.'), 3)) As [no3]
FROM (Select newid() AS [ID], JOBNUMBR from test) As [t]
GO

JOBNUMBR        | no1   | no2   | no3
:-------------- | :---- | :---- | :--
57299_87961_003 | 57299 | 87961 | 003
58OO7_88695_OO1 | 58OO7 | 88695 | OO1
66515_88567_OOO | 66515 | 88567 | OOO

db<>fiddle here

Answer 2

If your data is XML compliant, you can get a little more flexible by making it look like XML, then getting the data from there:

WITH V AS (
    SELECT * FROM (
        VALUES 
            ('123_456_789')
            ,('123_ABC_789')
            ,('1_4_7')

    ) [Values](ToSplit)

)
SELECT 
    q.x.value('d[1]', 'varchar(5)') A ,   
    q.x.value('d[2]', 'varchar(5)') B ,   
    q.x.value('d[3]', 'varchar(5)') C  
FROM V
    CROSS APPLY (
        SELECT CAST(CONCAT('<d>', REPLACE(ToSplit, '_', '</d><d>' ), '</d>') AS XML) x
    )  Q

OR

WITH V AS (
    SELECT * FROM (
        VALUES 
            ('123_456_789')
            ,('123_ABC_789')
            ,('1_4_7')

    ) [Values](ToSplit)

),
 Q as (
    SELECT CAST(CONCAT('<d>', REPLACE(ToSplit, '_', '</d><d>' ), '</d>') AS XML) x FROM V

 )
SELECT 
    x.value('d[1]', 'varchar(5)') A ,   
    x.value('d[2]', 'varchar(5)') B ,   
    x.value('d[3]', 'varchar(5)') C  
FROM Q

Both result in

A     B     C
----- ----- -----
123   456   789
123   ABC   789
1     4     7

The CTE V is just to setup the test data. You start by wrapping you data in a starting and ending tag, then replace your delimiter with an ending + starting tag. Convert all of this to XML, then get the value of the node ordinally.