February 16, 2010 at 10:36 am
Comments posted to this topic are about the item Retrieve numbers from a string
[font="Courier New"]____________________________________________________________________________________________
Remember as you walk down lifes road, don't forget to stop and pee on the bushes - Thordog[/font]
March 4, 2010 at 7:47 am
I noticed if the number was at the end of a sentence, it would not work. For example 'this is 9.00.', the period would mess it up. So I added a check for that right before the check for more than 1 decimal:
if right(@charnum,1) = '.' begin
set @charnum = left(@charnum,len(@charnum)-1)
end
-- don't convert numbers with more than 1 decimal
if len(replace(@charnum,'.','. ')) - len(@charnum) <= 1
begin
-- convert to float, insert into table
insert into @numbers values (@cnt,convert(float,@charnum))
end
That seemed to fix most of it, but I still have a problem with it now converting the string '9.00.' into 90. I have not had time to work on this again, so if you know a solution, I would be thankful!
[font="Courier New"]____________________________________________________________________________________________
Remember as you walk down lifes road, don't forget to stop and pee on the bushes - Thordog[/font]
March 8, 2010 at 4:59 am
Was looking at this and thinking there has to be a set-based way to acheive the same thing.
This is what I've come up with so far. Might be useful to read Jeff Modens article on Tally tables first (link in my sig below) if you're not already familiar with it.
I've omitted a few things on purpose so as not to obscure the main details of the query, like casting to decimal and checking for invalid strings with ISNUMERIC.
DECLARE @t TABLE(test varchar(8000))
INSERT @t
SELECT ' 1 iop 2?,.3e[]-4-@135.678;0--0;0...0'
SELECT
N,
number = SUBSTRING(test,N,PATINDEX('%[-.0123456789][^.0123456789]%',SUBSTRING(test+',',N,LEN(test)-N+2)))
FROM
@t, Tally
WHERE
N <= LEN(test) -- Restrict rows to the number of cahracters in the string
AND
CHARINDEX(SUBSTRING(test,N,1),'-.0123456789') > 0 -- Only rows starting with a valid numeric start character
AND
(N=0 OR CHARINDEX(SUBSTRING(test,N-1,1),'-.0123456789') = 0) -- either at the start of the string or following an invalid character.
This is basically an extrapolation of Jeffs string-splitting method.
In some basic tests I've done it's outperformed the looping method by a factor of 7 or 8.
I'm sure there are even faster methods and I've no doubt some of the more experienced posters on this site will be along to show us some CLR or XML method that's way beyond me 🙂
March 9, 2010 at 2:53 pm
Thanks Nigel, I got to play with this a little today. I have read Jeff's Tally table article before, good stuff! I knew there had to be a good set based way to do it. I will try and tweak it as I have time.
Thank you again 😀
[font="Courier New"]____________________________________________________________________________________________
Remember as you walk down lifes road, don't forget to stop and pee on the bushes - Thordog[/font]
March 10, 2010 at 1:52 am
No problem, thanks for the feedback.
Hope other people find it useful too.
May 12, 2016 at 7:31 am
Thanks for the script.
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply