I'm trying to write a UDF to translate a string that is either a guid or a project code associated with that guid into the guid:
CREATE FUNCTION fn_user_GetProjectID
(
#Project nvarchar(50)
)
RETURNS uniqueidentifier
AS
BEGIN
declare #ProjectID uniqueidentifier
BEGIN TRY
set #ProjectID = cast(#Project as uniqueidentifier)
END TRY
BEGIN CATCH
set #ProjectID = null
END CATCH
if(#ProjectID is null)
BEGIN
select #ProjectID = ProjectID from Project where projectcode = #Project
END
return #ProjectID
END
This works fine if the above code is embedded in my Stored Procedures, but I'd like to make a function out of it so that I follow DRY.
When I try to create the Function, I get errors like this:
Msg 443, Level 16, State 14, Procedure fn_user_GetProjectID, Line 16
Invalid use of side-effecting or time-dependent operator in 'BEGIN TRY' within a function.
Does anyone have an idea how I can get around this error?
Edit: I know I can't use Try-Catch in a Function, I guess a simplified questions would be, is there a way to do a cast that will just return NULL if the cast fails, instead of an error?
Apparently you can't use TRY-CATCH in a UDF.
According to this bug-reporting page for SQL Server:
Books Online documents this behaviour,
in topic "CREATE FUNCTION
(Transact-SQL)": "The following
statements are valid in a function:
[...] Control-of-Flow statements
except TRY...CATCH statements. [...]"
But they were giving hope for the future back in 2006:
However, this is a severe limitation
that should be removed in a future
release. You should post a suggestion
in this regard and I will
wholeheartedly vote for it.
From MSDN:
A column or local variable of
uniqueidentifier data type can be
initialized to a value in the
following ways:
By using the NEWID function.
By converting from a string constant
in the form
xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx,
in which each x is a hexadecimal digit
in the range 0-9 or a-f.
For example,
6F9619FF-8B86-D011-B42D-00C04FC964FF
is a valid uniqueidentifier value.
You can use pattern matching to verify the string. Note that this won't work for specific encoding that reduces the size of the GUID:
declare #Project nvarchar(50)
declare #ProjectID uniqueidentifier
declare #HexPattern nvarchar(268)
set #HexPattern =
'[A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9]' +
'[A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9]' +
'[A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9]' +
'[A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9][A-F0-9]'
/* Take into account GUID can have curly-brackets or be missing dashes */
/* Note: this will not work for GUIDs that have been specially encoded */
set #Project = '{' + CAST(NEWID() AS VARCHAR(36)) + '}'
select #Project
set #Project = REPLACE(REPLACE(REPLACE(#Project,'{',''),'}',''),'-','')
/* Cast as uniqueid if pattern matches, otherwise return null */
if #Project LIKE #HexPattern
select #ProjectID = CAST(
SUBSTRING(#Project,1,8) + '-' +
SUBSTRING(#Project,9,4) + '-' +
SUBSTRING(#Project,13,4) + '-' +
SUBSTRING(#Project,17,4) + '-' +
SUBSTRING(#Project,21,LEN(#Project)-20)
AS uniqueidentifier)
select #ProjectID
I know I can't use Try-Catch in a Function, I guess a simplified questions would be, is there a way to do a cast that will just return NULL if the cast fails, instead of an error?
Starting from SQL Server 2012 you could use TRY_CAST/TRY_CONVERT functions:
Returns a value cast to the specified data type if the cast succeeds; otherwise, returns null.
CREATE FUNCTION fn_user_GetProjectID(#Project nvarchar(50))
RETURNS uniqueidentifier
AS
BEGIN
declare #ProjectID uniqueidentifier = TRY_CAST(#Project as uniqueidentifier);
IF(#ProjectID is null)
BEGIN
select #ProjectID = ProjectID from Project where projectcode = #Project;
END
return #ProjectID;
END
Not sure, but why not flip it around... at first glance I would simplify it like this:
select #ProjectID =
ISNULL((select ProjectID from Project where
projectcode = #Project)
,(cast #Project as uniqueidentifier))
If this doesn't provide enough error handling, I'm sure there's a better way to pre-check that the cast can work without using try/catch...
My brute force method was to create my own ToGuid() function that verifies it can be converted to a GUID first, if not, it returns null. It may not be very fast but it does the job, and it is probably faster to convert the guid if it is one than to try to look it up in the table. EDIT: I meant to give credit to this blog, where I got the basis of my code for this function: http://jesschadwick.blogspot.com/2007/11/safe-handling-of-uniqueidentifier-in.html
CREATE FUNCTION [dbo].[ToGuid]
(
#input NVARCHAR(MAX)
)
RETURNS uniqueidentifier
AS
BEGIN
DECLARE #isValidGuid BIT;
DECLARE #temp NVARCHAR(MAX);
SET #isValidGuid = 1;
SET #temp = UPPER(LTRIM(RTRIM(REPLACE(REPLACE(REPLACE(#input, '-', ''), '{', ''), '}', ''))));
IF(#temp IS NOT NULL AND LEN(#temp) = 32)
BEGIN
DECLARE #index INT;
SET #index = 1
WHILE (#index <= 32)
BEGIN
IF (SUBSTRING(#temp, #index, 1) IN ('0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '0', 'A', 'B', 'C', 'D', 'E', 'F'))
BEGIN
SET #index = #index + 1
END
ELSE
BEGIN
SET #isValidGuid = 0
BREAK;
END
END
END
ELSE
BEGIN
SET #isValidGuid = 0
END
DECLARE #ret UNIQUEIDENTIFIER
IF(#isValidGuid = 1)
set #ret = cast(#input AS UNIQUEIDENTIFIER)
ELSE
set #ret = NULL
RETURN #ret
END
I'm still very interested if there is a better answer than this.
Verify if #Project is a number using the ISNUMERIC function.
your code should looks like that:
declare #ProjectID uniqueidentifier
set #ProjectID = null
IF ISNUMERIC(#Project) > 0
BEGIN
set #ProjectID = cast(#Project as uniqueidentifier)
END
if(#ProjectID is null)
BEGIN
select #ProjectID = ProjectID from Project where projectcode = #Project
END
return #ProjectID
I'm looking for a built-in function/extended function in T-SQL for string manipulation similar to the String.Format method in .NET.
If you are using SQL Server 2012 and above, you can use FORMATMESSAGE. eg.
DECLARE #s NVARCHAR(50) = 'World';
DECLARE #d INT = 123;
SELECT FORMATMESSAGE('Hello %s, %d', #s, #d)
-- RETURNS 'Hello World, 123'
More examples from MSDN: FORMATMESSAGE
SELECT FORMATMESSAGE('Signed int %i, %d %i, %d, %+i, %+d, %+i, %+d', 5, -5, 50, -50, -11, -11, 11, 11);
SELECT FORMATMESSAGE('Signed int with leading zero %020i', 5);
SELECT FORMATMESSAGE('Signed int with leading zero 0 %020i', -55);
SELECT FORMATMESSAGE('Unsigned int %u, %u', 50, -50);
SELECT FORMATMESSAGE('Unsigned octal %o, %o', 50, -50);
SELECT FORMATMESSAGE('Unsigned hexadecimal %x, %X, %X, %X, %x', 11, 11, -11, 50, -50);
SELECT FORMATMESSAGE('Unsigned octal with prefix: %#o, %#o', 50, -50);
SELECT FORMATMESSAGE('Unsigned hexadecimal with prefix: %#x, %#X, %#X, %X, %x', 11, 11, -11, 50, -50);
SELECT FORMATMESSAGE('Hello %s!', 'TEST');
SELECT FORMATMESSAGE('Hello %20s!', 'TEST');
SELECT FORMATMESSAGE('Hello %-20s!', 'TEST');
SELECT FORMATMESSAGE('Hello %20s!', 'TEST');
NOTES:
Undocumented in 2012
Limited to 2044 characters
To escape the % sign, you need to double it.
If you are logging errors in extended events, calling FORMATMESSAGE comes up as a (harmless) error
take a look at xp_sprintf. example below.
DECLARE #ret_string varchar (255)
EXEC xp_sprintf #ret_string OUTPUT,
'INSERT INTO %s VALUES (%s, %s)', 'table1', '1', '2'
PRINT #ret_string
Result looks like this:
INSERT INTO table1 VALUES (1, 2)
Just found an issue with the max size (255 char limit) of the string with this so there is an alternative function you can use:
create function dbo.fnSprintf (#s varchar(MAX),
#params varchar(MAX), #separator char(1) = ',')
returns varchar(MAX)
as
begin
declare #p varchar(MAX)
declare #paramlen int
set #params = #params + #separator
set #paramlen = len(#params)
while not #params = ''
begin
set #p = left(#params+#separator, charindex(#separator, #params)-1)
set #s = STUFF(#s, charindex('%s', #s), 2, #p)
set #params = substring(#params, len(#p)+2, #paramlen)
end
return #s
end
To get the same result as above you call the function as follows:
print dbo.fnSprintf('INSERT INTO %s VALUES (%s, %s)', 'table1,1,2', default)
I have created a user defined function to mimic the string.format functionality.
You can use it.
stringformat-in-sql
UPDATE:
This version allows the user to change the delimitter.
-- DROP function will loose the security settings.
IF object_id('[dbo].[svfn_FormatString]') IS NOT NULL
DROP FUNCTION [dbo].[svfn_FormatString]
GO
CREATE FUNCTION [dbo].[svfn_FormatString]
(
#Format NVARCHAR(4000),
#Parameters NVARCHAR(4000),
#Delimiter CHAR(1) = ','
)
RETURNS NVARCHAR(MAX)
AS
BEGIN
/*
Name: [dbo].[svfn_FormatString]
Creation Date: 12/18/2020
Purpose: Returns the formatted string (Just like in C-Sharp)
Input Parameters: #Format = The string to be Formatted
#Parameters = The comma separated list of parameters
#Delimiter = The delimitter to be used in the formatting process
Format: #Format = N'Hi {0}, Welcome to our site {1}. Thank you {0}'
#Parameters = N'Karthik,google.com'
#Delimiter = ','
Examples:
SELECT dbo.svfn_FormatString(N'Hi {0}, Welcome to our site {1}. Thank you {0}', N'Karthik,google.com', default)
SELECT dbo.svfn_FormatString(N'Hi {0}, Welcome to our site {1}. Thank you {0}', N'Karthik;google.com', ';')
*/
DECLARE #Message NVARCHAR(400)
DECLARE #ParamTable TABLE ( Id INT IDENTITY(0,1), Paramter VARCHAR(1000))
SELECT #Message = #Format
;WITH CTE (StartPos, EndPos) AS
(
SELECT 1, CHARINDEX(#Delimiter, #Parameters)
UNION ALL
SELECT EndPos + (LEN(#Delimiter)), CHARINDEX(#Delimiter, #Parameters, EndPos + (LEN(#Delimiter)))
FROM CTE
WHERE EndPos > 0
)
INSERT INTO #ParamTable ( Paramter )
SELECT
[Id] = SUBSTRING(#Parameters, StartPos, CASE WHEN EndPos > 0 THEN EndPos - StartPos ELSE 4000 END )
FROM CTE
UPDATE #ParamTable
SET
#Message = REPLACE(#Message, '{'+ CONVERT(VARCHAR, Id) + '}', Paramter )
RETURN #Message
END
There is a way, but it has its limitations. You can use the FORMATMESSAGE() function. It allows you to format a string using formatting similar to the printf() function in C.
However, the biggest limitation is that it will only work with messages in the sys.messages table. Here's an article about it: microsoft_library_ms186788
It's kind of a shame there isn't an easier way to do this, because there are times when you want to format a string/varchar in the database. Hopefully you are only looking to format a string in a standard way and can use the sys.messages table.
Coincidentally, you could also use the RAISERROR() function with a very low severity, the documentation for raiseerror even mentions doing this, but the results are only printed. So you wouldn't be able to do anything with the resulting value (from what I understand).
Good luck!
Raw t-sql is limited to CHARINDEX(), PATINDEX(), REPLACE(), and SUBSTRING() for string manipulation. But with sql server 2005 and later you can set up user defined functions that run in .Net, which means setting up a string.format() UDF shouldn't be too tough.
I think there is small correction while calculating end position.
Here is correct function
**>>**IF OBJECT_ID( N'[dbo].[FormatString]', 'FN' ) IS NOT NULL
DROP FUNCTION [dbo].[FormatString]
GO
/***************************************************
Object Name : FormatString
Purpose : Returns the formatted string.
Original Author : Karthik D V http://stringformat-in-sql.blogspot.com/
Sample Call:
SELECT dbo.FormatString ( N'Format {0} {1} {2} {0}', N'1,2,3' )
*******************************************/
CREATE FUNCTION [dbo].[FormatString](
#Format NVARCHAR(4000) ,
#Parameters NVARCHAR(4000)
)
RETURNS NVARCHAR(4000)
AS
BEGIN
--DECLARE #Format NVARCHAR(4000), #Parameters NVARCHAR(4000) select #format='{0}{1}', #Parameters='hello,world'
DECLARE #Message NVARCHAR(400), #Delimiter CHAR(1)
DECLARE #ParamTable TABLE ( ID INT IDENTITY(0,1), Parameter VARCHAR(1000) )
Declare #startPos int, #endPos int
SELECT #Message = #Format, #Delimiter = ','**>>**
--handle first parameter
set #endPos=CHARINDEX(#Delimiter,#Parameters)
if (#endPos=0 and #Parameters is not null) --there is only one parameter
insert into #ParamTable (Parameter) values(#Parameters)
else begin
insert into #ParamTable (Parameter) select substring(#Parameters,0,#endPos)
end
while #endPos>0
Begin
--insert a row for each parameter in the
set #startPos = #endPos + LEN(#Delimiter)
set #endPos = CHARINDEX(#Delimiter,#Parameters, #startPos)
if (#endPos>0)
insert into #ParamTable (Parameter)
select substring(#Parameters,#startPos,#endPos - #startPos)
else
insert into #ParamTable (Parameter)
select substring(#Parameters,#startPos,4000)
End
UPDATE #ParamTable SET #Message =
REPLACE ( #Message, '{'+CONVERT(VARCHAR,ID) + '}', Parameter )
RETURN #Message
END
Go
grant execute,references on dbo.formatString to public
One more idea.
Although this is not a universal solution - it is simple and works, at least for me :)
For one placeholder {0}:
create function dbo.Format1
(
#String nvarchar(4000),
#Param0 sql_variant
)
returns nvarchar(4000)
as
begin
declare #Null nvarchar(4) = N'NULL';
return replace(#String, N'{0}', cast(isnull(#Param0, #Null) as nvarchar(4000)));
end
For two placeholders {0} and {1}:
create function dbo.Format2
(
#String nvarchar(4000),
#Param0 sql_variant,
#Param1 sql_variant
)
returns nvarchar(4000)
as
begin
declare #Null nvarchar(4) = N'NULL';
set #String = replace(#String, N'{0}', cast(isnull(#Param0, #Null) as nvarchar(4000)));
return replace(#String, N'{1}', cast(isnull(#Param1, #Null) as nvarchar(4000)));
end
For three placeholders {0}, {1} and {2}:
create function dbo.Format3
(
#String nvarchar(4000),
#Param0 sql_variant,
#Param1 sql_variant,
#Param2 sql_variant
)
returns nvarchar(4000)
as
begin
declare #Null nvarchar(4) = N'NULL';
set #String = replace(#String, N'{0}', cast(isnull(#Param0, #Null) as nvarchar(4000)));
set #String = replace(#String, N'{1}', cast(isnull(#Param1, #Null) as nvarchar(4000)));
return replace(#String, N'{2}', cast(isnull(#Param2, #Null) as nvarchar(4000)));
end
and so on...
Such an approach allows us to use these functions in SELECT statement and with parameters of nvarchar, number, bit and datetime datatypes.
For example:
declare #Param0 nvarchar(10) = N'IPSUM' ,
#Param1 int = 1234567 ,
#Param2 datetime2(0) = getdate();
select dbo.Format3(N'Lorem {0} dolor, {1} elit at {2}', #Param0, #Param1, #Param2);
Actually there is no built in function similar to string.Format function of .NET is available in SQL server.
There is a function FORMATMESSAGE() in SQL server but it mimics to printf() function of C not string.Format function of .NET.
SELECT FORMATMESSAGE('This is the %s and this is the %s.', 'first variable', 'second variable') AS Result
Here is my version. Can be extended to accommodate more number of parameters and can extend formatting based on type. Currently only date and datetime types are formatted.
Example:
select dbo.FormatString('some string %s some int %s date %s','"abcd"',100,cast(getdate() as date),DEFAULT,DEFAULT)
select dbo.FormatString('some string %s some int %s date time %s','"abcd"',100,getdate(),DEFAULT,DEFAULT)
Output:
some string "abcd" some int 100 date 29-Apr-2017
some string "abcd" some int 100 date time 29-Apr-2017 19:40
Functions:
create function dbo.FormatValue(#param sql_variant)
returns nvarchar(100)
begin
/*
Tejasvi Hegde, 29-April-2017
Can extend formatting here.
*/
declare #result nvarchar(100)
if (SQL_VARIANT_PROPERTY(#param,'BaseType') in ('date'))
begin
select #result = REPLACE(CONVERT(CHAR(11), #param, 106), ' ', '-')
end
else if (SQL_VARIANT_PROPERTY(#param,'BaseType') in ('datetime','datetime2'))
begin
select #result = REPLACE(CONVERT(CHAR(11), #param, 106), ' ', '-')+' '+CONVERT(VARCHAR(5),#param,108)
end
else
begin
select #result = cast(#param as nvarchar(100))
end
return #result
/*
BaseType:
bigint
binary
char
date
datetime
datetime2
datetimeoffset
decimal
float
int
money
nchar
numeric
nvarchar
real
smalldatetime
smallint
smallmoney
time
tinyint
uniqueidentifier
varbinary
varchar
*/
end;
create function dbo.FormatString(
#format nvarchar(4000)
,#param1 sql_variant = null
,#param2 sql_variant = null
,#param3 sql_variant = null
,#param4 sql_variant = null
,#param5 sql_variant = null
)
returns nvarchar(4000)
begin
/*
Tejasvi Hegde, 29-April-2017
select dbo.FormatString('some string value %s some int %s date %s','"abcd"',100,cast(getdate() as date),DEFAULT,DEFAULT)
select dbo.FormatString('some string value %s some int %s date time %s','"abcd"',100,getdate(),DEFAULT,DEFAULT)
*/
declare #result nvarchar(4000)
select #param1 = dbo.formatValue(#param1)
,#param2 = dbo.formatValue(#param2)
,#param3 = dbo.formatValue(#param3)
,#param4 = dbo.formatValue(#param4)
,#param5 = dbo.formatValue(#param5)
select #param2 = cast(#param2 as nvarchar)
EXEC xp_sprintf #result OUTPUT,#format , #param1, #param2, #param3, #param4, #param5
return #result
end;
here's what I found with my experiments using the built-in
FORMATMESSAGE() function
sp_addmessage #msgnum=50001,#severity=1,#msgText='Hello %s you are #%d',#replace='replace'
SELECT FORMATMESSAGE(50001, 'Table1', 5)
when you call up sp_addmessage, your message template gets stored into the system table master.dbo.sysmessages (verified on SQLServer 2000).
You must manage addition and removal of template strings from the table yourself, which is awkward if all you really want is output a quick message to the results screen.
The solution provided by Kathik DV, looks interesting but doesn't work with SQL Server 2000, so i altered it a bit, and this version should work with all versions of SQL Server:
IF OBJECT_ID( N'[dbo].[FormatString]', 'FN' ) IS NOT NULL
DROP FUNCTION [dbo].[FormatString]
GO
/***************************************************
Object Name : FormatString
Purpose : Returns the formatted string.
Original Author : Karthik D V http://stringformat-in-sql.blogspot.com/
Sample Call:
SELECT dbo.FormatString ( N'Format {0} {1} {2} {0}', N'1,2,3' )
*******************************************/
CREATE FUNCTION [dbo].[FormatString](
#Format NVARCHAR(4000) ,
#Parameters NVARCHAR(4000)
)
RETURNS NVARCHAR(4000)
AS
BEGIN
--DECLARE #Format NVARCHAR(4000), #Parameters NVARCHAR(4000) select #format='{0}{1}', #Parameters='hello,world'
DECLARE #Message NVARCHAR(400), #Delimiter CHAR(1)
DECLARE #ParamTable TABLE ( ID INT IDENTITY(0,1), Parameter VARCHAR(1000) )
Declare #startPos int, #endPos int
SELECT #Message = #Format, #Delimiter = ','
--handle first parameter
set #endPos=CHARINDEX(#Delimiter,#Parameters)
if (#endPos=0 and #Parameters is not null) --there is only one parameter
insert into #ParamTable (Parameter) values(#Parameters)
else begin
insert into #ParamTable (Parameter) select substring(#Parameters,0,#endPos)
end
while #endPos>0
Begin
--insert a row for each parameter in the
set #startPos = #endPos + LEN(#Delimiter)
set #endPos = CHARINDEX(#Delimiter,#Parameters, #startPos)
if (#endPos>0)
insert into #ParamTable (Parameter) select substring(#Parameters,#startPos,#endPos)
else
insert into #ParamTable (Parameter) select substring(#Parameters,#startPos,4000)
End
UPDATE #ParamTable SET #Message = REPLACE ( #Message, '{'+CONVERT(VARCHAR,ID) + '}', Parameter )
RETURN #Message
END
Go
grant execute,references on dbo.formatString to public
Usage:
print dbo.formatString('hello {0}... you are {1}','world,good')
--result: hello world... you are good
At the moment this doesn't really exist (although you can of course write your own). There is an open connect bug for it: https://connect.microsoft.com/SQLServer/Feedback/Details/3130221, which as of this writing has just 1 vote.
Not exactly, but I would check out some of the articles on string handling (amongst other things) by "Phil Factor" (geddit?) on Simple Talk.
this is bad approach. you should work with assembly dll's, in which will do the same for you with better performance.