SQL: How to split column by character count [duplicate] - sql

Question

This question already has answers here:
Split column into multiple columns based on character count
(3 answers)
Closed last year.
I have one column with letters. I want to split this column into chunks of three. What SQL code for Microsoft would I need? I have read 'split my a special character' but I am not sure how to create a split by value where the split is not restricted to number of columns either.

Answer 1

You can do :
select t.*, substring(col, 1, 3), substring(col, 4, 3), substring(col, 7, 3)
from table t

Answer 2

If you really want to do this dynamically, as stated in the question, and have a query that creates just as many columns as needed, then you do need dynamic SQL.
Here is a solution that uses a recusive CTE to generate the query string.
declare #sql nvarchar(max);
with cte as (
select
1 pos,
cast('substring(code, 1, 3) col1' as nvarchar(max)) q,
max(len(code)) max_pos from mytable
union all
select
pos + 1,
cast(
q
+ ', substring(code, ' + cast(pos * 3 + 1 as nvarchar(3))
+ ', 3) col'
+ cast(pos + 1 as nvarchar(3))
as nvarchar(max)),
max_pos
from cte
where pos < max_pos / 3
)
select #sql = N'select ' + q + ' from mytable'
from cte
where len(q) = (select max(len(q)) from cte);
select #sql sql;
EXEC sp_executesql #sql;
The anchor of the recursive query computes the length of the longest string in column code. Then, the recursive part generates a series of substring() expressions for each chunk of 3 characters, with dynamic column names like col1, col2 and so on. You can then (debug and) execute that query string.
Demo on DB Fiddle:
-- debug
| sql |
| :---------------------------------------------------------------------------------------------------------------------------------- |
| select substring(code, 1, 3) col1, substring(code, 4, 3) col2, substring(code, 7, 3) col3, substring(code, 10, 3) col4 from mytable |
-- results
col1 | col2 | col3 | col4
:--- | :--- | :--- | :---
ABC | DEF | GHI |
XYZ | ABC | |
JKL | MNO | PQR | STU
ABC | DEF | |

Answer 3

Try it like this, which does not need any generic SQL (as long as you can specify a maximum count of columns):
First we need to define a mockup scenario to simulate your issue
DECLARE #tbl TABLE(ID INT IDENTITY, YourString VARCHAR(100));
INSERT INTO #tbl VALUES ('AB')
,('ABC')
,('ABCDEFGHI')
,('XYZABC')
,('JKLMNOPQRSTU')
,('ABCDEF');
--We can set the chunk length generically. Try it with other values...
DECLARE #ChunkLength INT=3;
--The query
SELECT p.*
FROM
(
SELECT t.ID
,CONCAT('Col',A.Nmbr) AS ColumnName
,SUBSTRING(t.YourString,(A.Nmbr-1)*#ChunkLength + 1,#ChunkLength) AS Chunk
FROM #tbl t
CROSS APPLY
(
SELECT TOP((LEN(t.YourString)+(#ChunkLength-1))/#ChunkLength) ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) FROM master..spt_values
) A(Nmbr)
) src
PIVOT
(
MAX(Chunk) FOR ColumnName IN(Col1,Col2,Col3,Col4,Col5,Col6 /*add the maximum column count here*/)
) p;
The idea in short:
By using an APPLY call we can create a row-wise tally. This will return multiple rows per input string. The row count is defined by the computed TOP-clause.
We use the row-wise tally first to create a column Name and second as parameters in SUBSTRING().
Finally we can use PIVOT to return this as horizontal list.
One hint about generic result sets:
This might be kind of religion, but - at least in my point of view - I would prefer a fix resultset with a lot of empty columns, rather than a generically defined set. The consumer should know the result format in advance...
You might use exactly the same query as dynamically created SQL statement. The only thing you would need to change is the actual list of column names in the PIVOT's IN-clause.