Convert a SQL function into a stored procedure - sql-server

Question

I am having trouble converting an UDF into a stored procedure.
Here is what I've got: this is the stored procedure that calls the function (I am using it to search for and remove all UNICODE characters that are not between 32 and 126):
ALTER PROCEDURE [dbo].[spRemoveUNICODE]
#FieldList varchar(250) = '',
#Multiple int = 0,
#TableName varchar(100) = ''
AS
BEGIN
SET NOCOUNT ON;
DECLARE #SQL VARCHAR(MAX), #counter INT = 0
IF #Multiple > 0
BEGIN
DECLARE #Field VARCHAR(100)
SELECT splitdata
INTO #TempValue
FROM dbo.fnSplitString(#FieldList,',')
WHILE (SELECT COUNT(*) FROM #TempValue) >= 1
BEGIN
DECLARE #Column VARCHAR(100) = (SELECT TOP 1 splitdata FROM #TempValue)
SET #SQL = 'UPDATE ' + #TableName + ' SET ' + #Column + ' = dbo.RemoveNonASCII(' + #Column + ')'
EXEC (#SQL)
--print #SQL
SET #counter = #counter + 1
PRINT #column + ' was checked for ' + #counter + ' rows.'
DELETE FROM #TempValue
WHERE splitdata = #Column
END
END
ELSE IF #Multiple = 0
BEGIN
SET #SQL = 'UPDATE ' + #TableName + ' SET ' + #FieldList + ' = dbo.RemoveNonASCII(' + #FieldList + ')'
EXEC (#SQL)
--print #SQL
SET #counter = #counter + 1
PRINT #column + ' was checked for ' + #counter + ' rows.'
END
END
And here is the UDF that I created to help with the update (RemoveNonASCII):
ALTER FUNCTION [dbo].[RemoveNonASCII]
(#nstring nvarchar(max))
RETURNS varchar(max)
AS
BEGIN
-- Variables
DECLARE #Result varchar(max) = '',#nchar nvarchar(1), #position int
-- T-SQL statements to compute the return value
set #position = 1
while #position <= LEN(#nstring)
BEGIN
set #nchar = SUBSTRING(#nstring, #position, 1)
if UNICODE(#nchar) between 32 and 127
set #Result = #Result + #nchar
set #position = #position + 1
set #Result = REPLACE(#Result,'))','')
set #Result = REPLACE(#Result,'?','')
END
if (#Result = '')
set #Result = null
-- Return the result
RETURN #Result
END
I've been trying to convert it into a stored procedure. I want to track how many rows actually get updated when this is run. Right now it just says that all rows, however many I run this on, are updated. I want to know if say only half of them had bad characters. The stored procedure is already set up so that it tells me which column it is looking at, I want to include how many rows were updated. Here is what I've tried so far:
DECLARE #Result varchar(max) = '',#nchar nvarchar(1), #position int, #nstring nvarchar(max), #counter int = 0, #CountRows int = 0, #Length int
--select Notes from #Temp where Notes is not null order by Notes OFFSET #counter ROWS FETCH NEXT 1 ROWS ONLY
set #nstring = (select Notes from #Temp where Notes is not null order by Notes OFFSET #counter ROWS FETCH NEXT 1 ROWS ONLY)
set #Length = LEN(#nstring)
if #Length = 0 set #Length = 1
-- Add the T-SQL statements to compute the return value here
set #position = 1
while #position <= #Length
BEGIN
print #counter
print #CountRows
select #nstring
set #nchar = SUBSTRING(#nstring, #position, 1)
if UNICODE(#nchar) between 32 and 127
begin
print unicode(#nchar)
set #Result = #Result + #nchar
set #counter = #counter + 1
end
if UNICODE(#nchar) not between 32 and 127
begin
set #CountRows = #CountRows + 1
end
set #position = #position + 1
END
print 'Rows found with invalid UNICODE: ' + convert(varchar,#CountRows)
Right now I'm purposely creating a temp table and adding a bunch of notes and then adding in a bunch of invalid characters.
I created a list of 700+ Notes and then updated 2 of them with some invalid characters (outside the 32 - 127). There are a few that are null and a few that are not null, but that doesn't have anything in them. What happens is that I get 0 updates.
Rows found with invalid UNICODE: 0
Though it does see that the UNICODE for the one that it pulls is 32.
Obviously I'm missing something I just don't see what it is.

Answer 1

Here is a set based solution to handle your bulk replacements. Instead of a slow scalar function this is utilizing an inline table valued function. These are far faster than their scalar ancestors. I am using a tally table here. I keep this as a view on my system like this.
create View [dbo].[cteTally] as
WITH
E1(N) AS (select 1 from (values (1),(1),(1),(1),(1),(1),(1),(1),(1),(1))dt(n)),
E2(N) AS (SELECT 1 FROM E1 a, E1 b), --10E+2 or 100 rows
E4(N) AS (SELECT 1 FROM E2 a, E2 b), --10E+4 or 10,000 rows max
cteTally(N) AS
(
SELECT ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) FROM E4
)
select N from cteTally
If you are interested about tally tables here is an excellent article on the topic. http://www.sqlservercentral.com/articles/T-SQL/62867/
create function RemoveNonASCII
(
#SearchVal nvarchar(max)
) returns table as
RETURN
with MyValues as
(
select substring(#SearchVal, N, 1) as MyChar
, t.N
from cteTally t
where N <= len(#SearchVal)
and UNICODE(substring(#SearchVal, N, 1)) between 32 and 127
)
select distinct MyResult = STUFF((select MyChar + ''
from MyValues mv2
order by mv2.N
--for xml path('')), 1, 0, '')
FOR XML PATH(''),TYPE).value('.','NVARCHAR(MAX)'), 1, 0, '')
from MyValues mv
;
Now instead of being forced to call this every single row you can utilize cross apply. The performance benefit of just this portion of your original question should be pretty huge.
I also eluded to your string splitter also being a potential performance issue. Here is an excellent article with a number of very fast set based string splitters. http://sqlperformance.com/2012/07/t-sql-queries/split-strings
The last step here would be eliminate the first loop in your procedure. This can be done also but I am not entirely certain what your code is doing there. I will look closer and see what I can find out. In the meantime parse through this and feel free to ask questions about any parts you don't understand.

Answer 2

Here is what I've got working based on the great help from Sean Lange:
How I call the Stored Procedure:
exec spRemoveUNICODE #FieldList='Notes,Notes2,Notes3,Notes4,Notes5',#Multiple=1,#TableName='#Temp'
The #Temp table is created:
create table #Temp (ID int,Notes nvarchar(Max),Notes2 nvarchar(max),Notes3 nvarchar(max),Notes4 nvarchar(max),Notes5 nvarchar(max))
Then I fill it with comments from 5 fields from a couple of different tables that range in length from NULL to blank (but not null) to 5000 characters.
I then insert some random characters like this:
update #Temp
set Notes2 = SUBSTRING(Notes2,1,LEN(Notes2)/2) + N'￿㹊潮Ņ᯸ࢹᖈư㹨ƶ槹鎤⻄ƺ綐ڌ⸀ƺ삸)䀤ƍ샄)Ņᛡ鎤ꗘᖃᒨ쬵Ğᘍ鎤ᐜᏰ>֔υ赸Ƹ쳰డ촜)鉀௿촜)쮜)Ἡ屰山舰霡ࣆ 耏Аం畠Ư놐ᓜતᏛ֔Ꮫ֨Ꮫ꯼ᓜƒ 邰఍厰ఆ邰఍드)抉鎤듄)繟Ĺ띨)᯸ࢹ䮸ࣉ᯸ࢹ䮸ࣉ샰)ԌƏ￿