tsql - How to split symbol data field on temp table to 5 columns?

Question

I work on a SQL Server 2012 query, an I face an issue: I can't split temp table field symbol data to 5 columns.

How to do that please?

CREATE TABLE #TEMP
(
    id INT IDENTITY(1,1),
    SymbolData  NVARCHAR(50)
)

INSERT INTO #TEMP (SymbolData)
VALUES ('0Hz ~ 4.5kHz'), ('0Hz | 9kHz'),
       ('0V - 4.5vl'), ('0Hz . 4.5kHz')

SELECT * FROM #TEMP

How to divide column Symbol Data into 5 columns to be

valuebefore unitbefore symbole valueafter unitafter
----------------------------------------------------
0            Hz           ~    4.5         Hz 
0            Hz           |     9          kHz 
0            V            -    4.5         vl
0            Hz           .    4.5         kHz 

Every row in the temp table column SymbolData has a symbol or character like | or - etc..

I need to split that column SymbolData into 5 parts every part represent column:

• get value before symbol on column as value before
• get unit before symbol on column as unit before
• get symbol on column as symbol
• get value after symbol on column as value after
• get unit after symbol on column as unit after

Answer 1

The solution is easy:

1. split the values by s(space) preserving order of elements
2. pivot the result
3. extract only numbers or only letters from the specific column

In order to split the values you can use XML like this. In order to extract only numbers you can perform a a chain of REPLACEs removing all units. In order to remove the numbers and leave the text, you can use REPLACEs again.

In my environment, I am using a lot of SQL CLR functions and the solution looks like this:

SELECT PVT.id
      ,PVT.symbolData
      ,dbo.fn_Utils_RegexReplace ([0], '[^d+]', '') AS [valuebefore]
      ,dbo.fn_Utils_RegexReplace ([0], 'd+', '') AS [unitbefore]
      ,[1] AS [symbole]
      ,dbo.fn_Utils_RegexReplace ([2], '[^d+.]', '') AS [valueafter]
      ,dbo.fn_Utils_RegexReplace ([2], '[d+.]', '') AS [unitafter]
FROM #TEMP
CROSS APPLY dbo.fn_Utils_RegexSplitWithOrder (SymbolData, 's') RS
PIVOT
(
    MAX([value]) FOR [index] IN ([0], [1], [2])
) PVT
ORDER BY PVT.id;

You can check this answer to get such functions in your environment, too.

In your case, it will be easier and safer to use XML to split the data and replace to shape the results.