May 2, 2018 at 7:35 pm
Jeff Moden - Wednesday, May 2, 2018 4:47 PMI'm seriously missing it in the thread above. What "small change" did Luis make?
Not to worry, it's easy to miss. He just made one small alteration...
My original...FOR XML PATH(''), TYPE).value('.', 'varchar(8000)')
Luis' modification...FOR XML PATH(''), TYPE).value('./text()[1]', 'varchar(8000)')
May 2, 2018 at 8:51 pm
Jason A. Long - Wednesday, May 2, 2018 7:35 PMJeff Moden - Wednesday, May 2, 2018 4:47 PMI'm seriously missing it in the thread above. What "small change" did Luis make?Not to worry, it's easy to miss. He just made one small alteration...
My original...
FOR XML PATH(''), TYPE).value('.', 'varchar(8000)')Luis' modification...
FOR XML PATH(''), TYPE).value('./text()[1]', 'varchar(8000)')
Thanks. That does make for a better looking execution plan but, in practice, I've found no particular performance advantage. Maybe it was just for the one test I did when Wayne Sheffield came out with his article on the subject of using FOR XML PATH to do concatenation.
--Jeff Moden
Change is inevitable... Change for the better is not.
May 2, 2018 at 9:15 pm
Here is the first round of testing...
Since I'm running 2017 on my home machine, I went ahead and threw in a STRING_AGG version as well...
To begin, the 3 current functions being tested...
CREATE FUNCTION dbo.tfn_ProperCase_JL
/* ===================================================================
05/02/2018 JL, Created:
This function will capitolize the first letter
in a string plusr any alpha character that follows a
non alpha character or apostrophe
=================================================================== */
--===== Define I/O parameters
(
@_string VARCHAR(8000)
)
RETURNS TABLE WITH SCHEMABINDING AS
RETURN
WITH
cte_n1 (n) AS (SELECT 1 FROM (VALUES (1),(1),(1),(1),(1),(1),(1),(1),(1),(1)) n (n)),
cte_n2 (n) AS (SELECT 1 FROM cte_n1 a CROSS JOIN cte_n1 b),
cte_Tally (n) AS (
SELECT TOP (LEN(@_string))
ROW_NUMBER() OVER (ORDER BY b.n)
FROM
cte_n2 a CROSS JOIN cte_n2 b
)
SELECT
CasedString = ((
SELECT
CONCAT('', cv.cased_value)
FROM
cte_Tally t
CROSS APPLY ( VALUES (
CASE
WHEN SUBSTRING(@_string, t.n - 1, 1) NOT LIKE '[''a-Z]'
THEN UPPER(SUBSTRING(@_string, t.n, 1))
ELSE LOWER(SUBSTRING(@_string, t.n, 1))
END
) ) cv (cased_value)
FOR XML PATH(''), TYPE).value('.', 'varchar(8000)')
);
GO
CREATE FUNCTION dbo.tfn_ProperCase_LC
/* ===================================================================
05/02/2018 JL, Created:
This is an exact copy of the dbo.tfn_ProperCase_JL
function except for a slight alteration in the FOR XML PATH line
that was suggested by Luis Cazares
=================================================================== */
--===== Define I/O parameters
(
@_string VARCHAR(8000)
)
RETURNS TABLE WITH SCHEMABINDING AS
RETURN
WITH
cte_n1 (n) AS (SELECT 1 FROM (VALUES (1),(1),(1),(1),(1),(1),(1),(1),(1),(1)) n (n)),
cte_n2 (n) AS (SELECT 1 FROM cte_n1 a CROSS JOIN cte_n1 b),
cte_Tally (n) AS (
SELECT TOP (LEN(@_string))
ROW_NUMBER() OVER (ORDER BY b.n)
FROM
cte_n2 a CROSS JOIN cte_n2 b
)
SELECT
CasedString = ((
SELECT
CONCAT('', cv.cased_value)
FROM
cte_Tally t
CROSS APPLY ( VALUES (
CASE
WHEN SUBSTRING(@_string, t.n - 1, 1) NOT LIKE '[''a-Z]'
THEN UPPER(SUBSTRING(@_string, t.n, 1))
ELSE LOWER(SUBSTRING(@_string, t.n, 1))
END
) ) cv (cased_value)
FOR XML PATH(''), TYPE).value('./text()[1]', 'varchar(8000)')
);
GO
CREATE FUNCTION dbo.tfn_ProperCase_SA
/* ===================================================================
05/02/2018 JL, Created:
This returns the same results as the dbo.tfn_ProperCase_JL
function but replaces the "FOR XML PATH" syntax with the
newer STRING_AGG function
=================================================================== */
--===== Define I/O parameters
(
@_string VARCHAR(8000)
)
RETURNS TABLE WITH SCHEMABINDING AS
RETURN
WITH
cte_n1 (n) AS (SELECT 1 FROM (VALUES (1),(1),(1),(1),(1),(1),(1),(1),(1),(1)) n (n)),
cte_n2 (n) AS (SELECT 1 FROM cte_n1 a CROSS JOIN cte_n1 b),
cte_Tally (n) AS (
SELECT TOP (LEN(@_string))
ROW_NUMBER() OVER (ORDER BY b.n)
FROM
cte_n2 a CROSS JOIN cte_n2 b
)
SELECT
CasedString = STRING_AGG(cv.cased_value, '')
FROM
cte_Tally t
CROSS APPLY ( VALUES (
CASE
WHEN SUBSTRING(@_string, t.n - 1, 1) NOT LIKE '[''a-Z]'
THEN UPPER(SUBSTRING(@_string, t.n, 1))
ELSE LOWER(SUBSTRING(@_string, t.n, 1))
END
) ) cv (cased_value);
GO
The test harness... (being executed with "Discard results after execution" turned on)
IF OBJECT_ID('tempdb..#TestData', 'U') IS NULL
BEGIN -- DROP TABLE #TestData;
CREATE TABLE #TestData (
rid INT NOT NULL PRIMARY KEY CLUSTERED,
string_val VARCHAR(1000)
);
WITH
cte_n1 (n) AS (SELECT 1 FROM (VALUES (1),(1),(1),(1),(1),(1),(1),(1),(1),(1)) n (n)),
cte_n2 (n) AS (SELECT 1 FROM cte_n1 a CROSS JOIN cte_n1 b),
cte_Tally (n) AS (
SELECT TOP (10000)
ROW_NUMBER() OVER (ORDER BY (SELECT NULL))
FROM
cte_n2 a CROSS JOIN cte_n2 b
)
INSERT #TestData(rid, string_val)
SELECT
t.n,
x.string_val
FROM
cte_Tally t
CROSS APPLY ( VALUES ((
SELECT TOP (ABS(CHECKSUM(NEWID())) % 999 + 1)
CONCAT('', CHAR(a2.ascii_val))
FROM
cte_Tally t2
CROSS APPLY ( VALUES (ABS(CHECKSUM(NEWID())) % 58 + 65) ) av (ascii_val)
CROSS APPLY ( VALUES (
CASE av.ascii_val
WHEN 91 THEN 32
WHEN 92 THEN 39
WHEN 93 THEN 44
WHEN 94 THEN 46
ELSE av.ascii_val
END)
) a2 (ascii_val)
WHERE
t.n > 0
FOR XML PATH(''), TYPE).value('./text()[1]', 'varchar(8000)')
) ) x (string_val);
END;
--================================================================================================================================================
--================================================================================================================================================
--================================================================================================================================================
SET NOCOUNT ON;
GO
DBCC FREEPROCCACHE WITH NO_INFOMSGS;
DBCC DROPCLEANBUFFERS WITH NO_INFOMSGS;
PRINT(CONCAT(CHAR(13), CHAR(10), N' wait a moment...', CHAR(13), CHAR(10)));
WAITFOR DELAY '00:00:01';
GO
-- SET STATISTICS XML ON;
GO
DECLARE @_clock_start DATETIME2(7) = SYSDATETIME(), @_test_name NVARCHAR(70) = N'tfn_ProperCase_JL ';
PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(CONCAT(N' Start Time: ', @_clock_start, N' Test Name: ', ISNULL(NULLIF(@_test_name, N' '), N'Test Query 1'),
REPLICATE(N' ', 100)), 148), N'¦', CHAR(13), CHAR(10), N'+', REPLICATE(N'-', 148), N'+')); SET STATISTICS IO ON;
-- ____________________________________________________________________________________________________________________________________________
-- ?????? place tsql here ????????????????????????????????????????????????????????????
SELECT
td.rid,
td.string_val,
pc.CasedString
FROM
#TestData td
CROSS APPLY dbo.tfn_ProperCase_JL(td.string_val) pc;
-- ?????? place tsql here ????????????????????????????????????????????????????????????
-- ????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
DECLARE @_clock_stop DATETIME2(7) = SYSDATETIME(); SET STATISTICS IO OFF; PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(STUFF(CONCAT(N' Finish Time: ', @_clock_stop, N' Duration: ', DATEDIFF(mcs, @_clock_start, @_clock_stop)
/ 1000000.0, N' secs. ', DATEDIFF(mcs, @_clock_start, @_clock_stop) / 1000.0, N' ms.', REPLICATE(N' ', 100)), 76, 3, N''), 148), N'¦
+', REPLICATE(N'-', 148), N'+'));
GO
-- SET STATISTICS XML OFF;
GO
DBCC FREEPROCCACHE WITH NO_INFOMSGS;
DBCC DROPCLEANBUFFERS WITH NO_INFOMSGS;
PRINT(CONCAT(CHAR(13), CHAR(10), N' wait a moment...', CHAR(13), CHAR(10)));
WAITFOR DELAY '00:00:01';
GO
-- SET STATISTICS XML ON;
GO
DECLARE @_clock_start DATETIME2(7) = SYSDATETIME(), @_test_name NVARCHAR(70) = N'tfn_ProperCase_LC ';
PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(CONCAT(N' Start Time: ', @_clock_start, N' Test Name: ', ISNULL(NULLIF(@_test_name, N' '), N'Test Query 2'),
REPLICATE(N' ', 100)), 148), N'¦', CHAR(13), CHAR(10), N'+', REPLICATE(N'-', 148), N'+')); SET STATISTICS IO ON;
-- ____________________________________________________________________________________________________________________________________________
-- ?????? place tsql here ????????????????????????????????????????????????????????????
SELECT
td.rid,
td.string_val,
pc.CasedString
FROM
#TestData td
CROSS APPLY dbo.tfn_ProperCase_LC(td.string_val) pc;
-- ?????? place tsql here ????????????????????????????????????????????????????????????
-- ????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
DECLARE @_clock_stop DATETIME2(7) = SYSDATETIME(); SET STATISTICS IO OFF; PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(STUFF(CONCAT(N' Finish Time: ', @_clock_stop, N' Duration: ', DATEDIFF(mcs, @_clock_start, @_clock_stop)
/ 1000000.0, N' secs. ', DATEDIFF(mcs, @_clock_start, @_clock_stop) / 1000.0, N' ms.', REPLICATE(N' ', 100)), 76, 3, N''), 148), N'¦
+', REPLICATE(N'-', 148), N'+'));
GO
-- SET STATISTICS XML OFF;
GO
DBCC FREEPROCCACHE WITH NO_INFOMSGS;
DBCC DROPCLEANBUFFERS WITH NO_INFOMSGS;
PRINT(CONCAT(CHAR(13), CHAR(10), N' wait a moment...', CHAR(13), CHAR(10)));
WAITFOR DELAY '00:00:01';
GO
-- SET STATISTICS XML ON;
GO
DECLARE @_clock_start DATETIME2(7) = SYSDATETIME(), @_test_name NVARCHAR(70) = N'tfn_ProperCase_SA ';
PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(CONCAT(N' Start Time: ', @_clock_start, N' Test Name: ', ISNULL(NULLIF(@_test_name, N' '), N'Test Query 2'),
REPLICATE(N' ', 100)), 148), N'¦', CHAR(13), CHAR(10), N'+', REPLICATE(N'-', 148), N'+')); SET STATISTICS IO ON;
-- ____________________________________________________________________________________________________________________________________________
-- ?????? place tsql here ????????????????????????????????????????????????????????????
SELECT
td.rid,
td.string_val,
pc.CasedString
FROM
#TestData td
CROSS APPLY dbo.tfn_ProperCase_SA(td.string_val) pc;
-- ?????? place tsql here ????????????????????????????????????????????????????????????
-- ????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
DECLARE @_clock_stop DATETIME2(7) = SYSDATETIME(); SET STATISTICS IO OFF; PRINT(CONCAT(N'+', REPLICATE(N'-', 148), N'+
¦', LEFT(STUFF(CONCAT(N' Finish Time: ', @_clock_stop, N' Duration: ', DATEDIFF(mcs, @_clock_start, @_clock_stop)
/ 1000000.0, N' secs. ', DATEDIFF(mcs, @_clock_start, @_clock_stop) / 1000.0, N' ms.', REPLICATE(N' ', 100)), 76, 3, N''), 148), N'¦
+', REPLICATE(N'-', 148), N'+'));
GO
-- SET STATISTICS XML OFF;
GO
The results...
wait a moment...
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Start Time: 2018-05-02 23:11:35.6157441 Test Name: tfn_ProperCase_JL
+----------------------------------------------------------------------------------------------------------------------------------------------------+
Table '#TestData___________________________________________________________________________________________________________000000000077'. Scan count 1, logical reads 676, physical reads 0, read-ahead reads 465, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Finish Time: 2018-05-02 23:11:38.2096084 Duration: 2.593864 secs. 2593.864000 ms.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
wait a moment...
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Start Time: 2018-05-02 23:11:39.2327236 Test Name: tfn_ProperCase_LC
+----------------------------------------------------------------------------------------------------------------------------------------------------+
Table '#TestData___________________________________________________________________________________________________________000000000077'. Scan count 1, logical reads 676, physical reads 0, read-ahead reads 465, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Finish Time: 2018-05-02 23:11:41.7875599 Duration: 2.554836 secs. 2554.836000 ms.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
wait a moment...
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Start Time: 2018-05-02 23:11:42.8179316 Test Name: tfn_ProperCase_SA
+----------------------------------------------------------------------------------------------------------------------------------------------------+
Table '#TestData___________________________________________________________________________________________________________000000000077'. Scan count 1, logical reads 676, physical reads 0, read-ahead reads 465, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
¦ Finish Time: 2018-05-02 23:11:44.7733370 Duration: 1.955406 secs. 1955.406000 ms.
+----------------------------------------------------------------------------------------------------------------------------------------------------+
Actual execution plan attached...
May 2, 2018 at 9:44 pm
Jeff Moden - Wednesday, May 2, 2018 8:51 PMJason A. Long - Wednesday, May 2, 2018 7:35 PMJeff Moden - Wednesday, May 2, 2018 4:47 PMI'm seriously missing it in the thread above. What "small change" did Luis make?Not to worry, it's easy to miss. He just made one small alteration...
My original...
FOR XML PATH(''), TYPE).value('.', 'varchar(8000)')Luis' modification...
FOR XML PATH(''), TYPE).value('./text()[1]', 'varchar(8000)')Thanks. That does make for a better looking execution plan but, in practice, I've found no particular performance advantage. Maybe it was just for the one test I did when Wayne Sheffield came out with his article on the subject of using FOR XML PATH to do concatenation.
I'll make the switch because a) the impact on estimated costing seems more in line with reality and b) based on the very limited testing I've done today, it does appear to be marginally faster, most of the time.
May 3, 2018 at 4:05 am
There is also a version that allows for exceptions such as names with apostrophes etc. here:
http://www.wisesoft.co.uk/scripts/tsql_proper_case_udf.aspx
P.S. I have no connection with the Author of the above function.
May 3, 2018 at 8:33 am
Jeff Moden - Wednesday, May 2, 2018 7:37 AMjonathan.crawford - Wednesday, May 2, 2018 6:14 AMJeff Moden - Initial creation and unit test (http://www.sqlservercentral.com/Forums/Topic530630-8-2.aspx)
Wow... that was a while back. I'm humbled that you kept a link to it. It also reminded me that I was going to try something a bit different so that it would work in an iTVF using a reverse order Tally table. I guess I never got back to it because I never needed to use such a thing.
I also like the Nested Replaces that Chris did alot. Should be nasty fast especially with the binary collation that was used.
I just pulled the link out of the comments in our system function that we stole from you. 🙂
-------------------------------------------------------------------------------------------------------------------------------------
Please follow Best Practices For Posting On Forums to receive quicker and higher quality responses
May 3, 2018 at 11:08 am
jonathan.crawford - Thursday, May 3, 2018 8:33 AMJeff Moden - Wednesday, May 2, 2018 7:37 AMjonathan.crawford - Wednesday, May 2, 2018 6:14 AMJeff Moden - Initial creation and unit test (http://www.sqlservercentral.com/Forums/Topic530630-8-2.aspx)
Wow... that was a while back. I'm humbled that you kept a link to it. It also reminded me that I was going to try something a bit different so that it would work in an iTVF using a reverse order Tally table. I guess I never got back to it because I never needed to use such a thing.
I also like the Nested Replaces that Chris did alot. Should be nasty fast especially with the binary collation that was used.
I just pulled the link out of the comments in our system function that we stole from you. 🙂
Heh... SEE???!!! Documentation works!
--Jeff Moden
Change is inevitable... Change for the better is not.
Viewing 7 posts - 16 through 21 (of 21 total)
You must be logged in to reply to this topic. Login to reply