How did you split strings before string_split()

Question

I have some strings I need to split on a SQL Server 2014 instance.

I would normally do it something like this:

declare @bk nvarchar(1000)

set @bk = 'SIS5.0~person_id~696969'

select value from string_split(@bk, '~')

and I can also do it like this:

declare @bk nvarchar(1000)

set @bk = 'SIS5.0~person_id~696969'

;with hizizzle as (
select left(@bk, charindex('~', @bk)-1) as flippity, right(@bk, len(@bk) - charindex('~', @bk)) as floppity)

select flippity, left(floppity, charindex('~',floppity)-1) as floppity, 
right(floppity, len(floppity) - charindex('~', floppity)) as flooo 
from hizizzle

...but there must be a better way, right?

Question: How did you do string splits in the ancient past before 2014?

Answer 1

If you know the number of items and want to access one or more by its index, you can do something like the following:

DECLARE @bk nvarchar(1000)

SET @bk = 'SIS5.0~person_id~696969'

SELECT
CAST(N'<x>' + REPLACE(@bk, '~', N'</x><x>') + N'</x>' AS XML).value('/x[1]', 'nvarchar(max)') [string1],
CAST(N'<x>' + REPLACE(@bk, '~', N'</x><x>') + N'</x>' AS XML).value('/x[2]', 'nvarchar(max)') [string2],
CAST(N'<x>' + REPLACE(@bk, '~', N'</x><x>') + N'</x>' AS XML).value('/x[3]', 'nvarchar(max)') [string3]


string1    string2    string3
---------- ---------- -------
SIS5.0     person_id  696969

Splits on the '~' character, surrounds with simple tags, then uses the .value property to get a specific index.

Works for 2008, and I believe (but can't remember) it also worked on 2005.

Answer 2

From comments:

1. The built-in STRING_SPLIT() function was added in SQL Server 2016, not 2014. And for me, I go the SQLCLR route because I am not using Azure Single DB or AWS RDS and thus have no reason not to. I use the SQLCLR library that I created, SQL#, which contains String_Split and String_SplitInts (optimized for splitting a list of BIGINT / INT / SMALLINT / TINYINT values). The "4k" versions of both are to be used when the input string is guaranteed to never be more than NVARCHAR(4000) as they are faster than the non-"4k" versions. – Solomon Rutzky

2. there were lots of User-Defined Table-Valued functions floating around that did this. Here's a survey article by Aaron Bertrand: sqlperformance.com/2012/07/t-sql-queries/split-strings – David Browne - Microsoft

3. this is probably the most efficient by Jeff Moden sqlservercentral.com/articles/… – Squirrel

Answer 3

declare @bk nvarchar(1000)

set @bk = 'SIS5.0~person_id~696969'

;with hizizzle as (
select left(@bk, charindex('~', @bk)-1) as flippity, right(@bk, len(@bk) - charindex('~', @bk)) as floppity)

select flippity, left(floppity, charindex('~',floppity)-1) as floppity, 
right(floppity, len(floppity) - charindex('~', floppity)) as flooo 
from hizizzle

It was pretty much this - it sucked!