Dear SQLers:

I have finished writing my first involved stored procedure. It
makes strings nice. It works, but it might be better.

Have I, for example, missed out on functions that could have made
it easier?

replace() does not work as I expected when the search string is
partly contained in the result of a replace
replace(N'three spaces',N' ',N' ')
will result in
N'three spaces'
not
N'three spaces'
That is why I loop.

I also do not know how SQL Server handles multi-line values. I
am assuming there is a CR (or UniCode equivalent) to delimit lines. In
case there could be LFs, I eliminate them in the line flagged
"--*****TRY".

Test code follows.

What do you think?

***** Start of Code *****
-- NiceString
-- This procedure makes a string "nice". "nice" means that each line
of the
-- string has:
-- 1) no occurrences of <tab> (each replaced with space), and
-- 2) no leading spaces (removed) (optional),
-- 3) no trailing spaces (removed) (optional),
-- 4) no two spaces in a row (squeezed to one space).
-- The four points are applied in order 1-4.

-- "cLines": count of lines in the string. This is defined as the
number of
-- line delimiters in the string + 1. e.g. (where <LD> is a line
delimiter):
-- N'' has one line.
-- N'kitty' has one line.
-- N'This is the first line,<LD>
-- and this is the last line.' has two lines.
-- N'This is the first line,<LD>
-- this is the middle line,<LD>
-- and this is the last line.' has three lines.

-- If mLines=0, returns cLines.
-- If mLines>0 and cLines<=mLines, returns cLines.
-- If mLines>0 and cLines>mLines, returns a value >mLines.

create procedure NiceString
@strNiceMe nvarchar(max) output, -- the string to make nice and the
result
@mLines int=0, -- maximum number of lines permitted in string
-- If 0, the string can have any number of lines. If >0 and the
string
-- turns out to have more lines, processing will be aborted and the
string
-- will not be changed.
@fTrimLead bit=1, -- trim the leading blanks on each line? 0: no,
1: yes
@fTrimTrail bit=1 -- trim the trailing blanks on each line? 0: no,
1: yes
with recompile
as
begin
declare @strWork nvarchar(max)=@strNiceMe -- working version of
string
declare @strOut nvarchar(max)=N'' -- result string so far
declare @cLines int=1 -- number of lines (always at least one)
declare @strLine nvarchar(max) -- the current line being processed
declare @chrLineDelim nchar(1)=nchar(13) -- CR
declare @chrLF nchar(1)=nchar(10) -- LF
declare @chrTAB nchar(1)=nchar(9) -- TAB

declare @iLineDelim int -- where next line delimiter is in
@strWork

declare @cLeading int -- how many leading spaces in line
declare @cTrailing int -- how many trailing spaces in line

declare @strPrev nvarchar(max) -- previous version of line

while len(@strWork)>0 and (@mLines=0 or @cLines<=@mLines)
begin
-- Get the next line and remove it from strWork.
select @iLineDelim=patindex(N'%'+@chrLineDelim+N'%',@strW ork)
if @iLineDelim>0
begin
select @strLine=left(@strWork,@iLineDelim-1)
select
@strWork=substring(@strWork,@iLineDelim+1,len(@str Work))
select @cLines=@cLines+1
end
else
begin
select @strLine=@strWork
select @strWork=N''
end

-- Trimming and removal of tabs
select @strLine=replace(@strLine,@chrLF,N'') --*****TRY
select @strLine=replace(@strLine,@chrTAB,N' ')
-- If either end is not to be trimmed, then the leading and
trailing
-- spaces will have to be restored after the line is squeezed.
if @fTrimLead=1 and @fTrimTrail=1
begin
select @cLeading=0
select @cTrailing=0
end
else
begin
select @cLeading=patindex(N'%[^ ]%',@strLine)
if @cLeading=len(@strLine) -- Line is only blanks.
select @cTrailing=0
else
select @cTrailing=patindex(N'%[^ ]%',reverse(@strLine))
end
select @strLine=ltrim(rtrim(@strLine))

-- Compression of consecutive spaces
select @strPrev=N''
while @strLine<>@strPrev
begin
select @strPrev=@strLine
select @strLine=replace(@strLine,N' ',N' ')
end

-- Add back any leading or trailing spaces being kept.
if @fTrimLead=0
select @strLine=space(@cLeading)+@strLine
if @fTrimTrail=0
select @strLine=@strLine+space(@cTrailing)

-- Add modified line to @strOut.
if len(@strOut)>0
select @strOut=@strOut+@chrLineDelim
select @strOut=@strOut+@strLine
end

if @mLines=0 or @cLines<=@mLines -- Replace original string if no
error.
select @strNiceMe=@strOut
return @cLines
end
***** End of Code *****

***** Start of Test Code *****
declare @strTestValue nvarchar(max)=
N' This is an example. '
declare @strTest nvarchar(max)
declare @RetVal int

select N'raw string',N'>'+@strTestValue+N'<'

select @strTest=@strTestValue
execute @RetVal=NiceString @strTest output
select 'defaults not given',@RetVal as N'Ret Val',N'>'+@strTest+N'<'

select @strTest=@strTestValue
execute @RetVal=NiceString @strTest output,0,1,1
select 'defaults given',@RetVal as N'Ret Val',N'>'+@strTest+N'<'

select @strTest=@strTestValue
execute @RetVal=NiceString @strTest output,0,0,0
select '0,0,0',@RetVal as N'Ret Val',N'>'+@strTest+N'<'

select @strTest=@strTestValue
execute @RetVal=NiceString @strTest output,0,1,0
select '0,1,0',@RetVal as N'Ret Val',N'>'+@strTest+N'<'

select @strTest=@strTestValue
execute @RetVal=NiceString @strTest output,0,0,1
select '0,0,1',@RetVal as N'Ret Val',N'>'+@strTest+N'<'

select @strTest=N''
execute @RetVal=NiceString @strTest output
select 'empty, defaults not given',@RetVal as N'Ret
Val',N'>'+@strTest+N'<'
select
len(@strTest) as Length,
patindex(N'%'+nchar(13)+N'%',@strTest) as N'First LD'

select @strTest=
N' This is a three line string. '+nchar(13)+
N' This is the second line. '+nchar(13)+
N' This is the last line. '
execute @RetVal=NiceString @strTest output
select 'defaults not given',@RetVal as N'Ret Val',N'>'+@strTest+N'<'
select
len(@strTest) as Length,
patindex(N'%'+nchar(13)+N'%',@strTest) as N'First LD'

select @strTest=
N' This is a three line string. '+nchar(13)+
N' This is the second line. '+nchar(13)+
N' This is the last line. '
execute @RetVal=NiceString @strTest output,3
select '3 rest default',@RetVal as N'Ret Val',N'>'+@strTest+N'<'
select
len(@strTest) as Length,
patindex(N'%'+nchar(13)+N'%',@strTest) as N'First LD'

select @strTest=
N' This is a three line string. '+nchar(13)+
N' This is the second line. '+nchar(13)+
N' This is the last line. '
execute @RetVal=NiceString @strTest output,2
select '2 rest default',@RetVal as N'Ret Val',N'>'+@strTest+N'<'
select
len(@strTest) as Length,
patindex(N'%'+nchar(13)+N'%',@strTest) as N'First LD'
***** End of Test Code *****

Sincerely,

Gene Wirchenko

Gene Wirchenko (genew (AT) ocis (DOT) net) writes:
Kind of. While you can write this stuff in T-SQL, this is exactly why
Microsoft added the capability to put code written in .Net in SQL Server.

SQL Server does not really care much about lines at all. If you put CR-LF at
the end of the string, there's a CR-LF at the end of the string. If string
only has LF, that's it. Or only CR. It all depends on where the data comes
from. If you load a text file from Unix, there will be only LF. If you
run examples from SSMS there will probably be CR-LF, since SSMS in a Windows
application.

I do not have the time to look at the code as such for now.

--
Erland Sommarskog, SQL Server MVP, esquel (AT) sommarskog (DOT) se

Books Online for SQL Server 2005 at
http://www.microsoft.com/technet/pro...ads/books.mspx
Books Online for SQL Server 2000 at
http://www.microsoft.com/sql/prodinf...ons/books.mspx

On Wed, 6 Apr 2011 07:43:02 +0000 (UTC), Erland Sommarskog
<esquel (AT) sommarskog (DOT) se> wrote:

I have not gotten into .Net. One at a time.

What I am going to put on top of this is an insert/update trigger
to clean up strings. Can I call .Net code from within a trigger?

Well, that could make things interesting in a cross-platform
situation.

SSMS does not display the multi-line values as such. There is
space between the lines as in:
LINE ONE line two

Sincerely,

Gene Wirchenko

On 05-04-2011 23:26, Gene Wirchenko wrote:
if there's only one statement to execute, see below.

Gene Wirchenko (genew (AT) ocis (DOT) net) writes:
Yes. I think you would put this in a function, and calling CLR UDF is just
the same as calling a T-SQL UDF.

You can even write triggers directly in the CLR. I will have that I have
yet to see a use case for that.

--
Erland Sommarskog, SQL Server MVP, esquel (AT) sommarskog (DOT) se

Links for SQL Server Books Online:
SQL 2008: http://msdn.microsoft.com/en-us/sqlserver/cc514207.aspx
SQL 2005: http://msdn.microsoft.com/en-us/sqlserver/bb895970.aspx

On Wed, 06 Apr 2011 21:15:48 +0200, Henk van den Berg
<hvandenberg (AT) xs4all (DOT) nl> wrote:

I do not like the clutter. To guard against the risk you imply
(and it is a valid concern), instead, I am VERY fussy with indentation
as in, in C:
for (i=1; i<=10000; i++)
; /* empty body */

Sincerely,

Gene Wirchenko

Gene Wirchenko (genew (AT) ocis (DOT) net) writes:

My personal standard is that I don't have BEGIN END, if it is a single
line, but if there are more than one line, I have BEGIN END, even if
it's a single statement.


--
Erland Sommarskog, SQL Server MVP, esquel (AT) sommarskog (DOT) se

Links for SQL Server Books Online:
SQL 2008: http://msdn.microsoft.com/en-us/sqlserver/cc514207.aspx
SQL 2005: http://msdn.microsoft.com/en-us/sqlserver/bb895970.aspx

I like the idea of white space replacement without loops presented
here:

http://www.itjungle.com/fhg/fhg101106-story02.html

replace(replace(replace(name,' ','<>'),'><',''),'<>',' ')

1) replace each white space with <>
2) when a white space follows an other, it looks like this: >< -
remove them
3) finally there is just <> left, replace it with a normal white space

brgds

Philipp Post