Re: SQL function equivalent to TestN?

"Likely performed better, yes. but with much less accuracy."

Accuracy is a vague term without being given in the context of the use
case(s), the data scenarios, and how much slop in the data you're trying to
accommodate, or not.

There are simple use cases and data scenarios where TRANSLATE gets the job
done equally well (only faster), and other times when the vastly more
capable, yet slower, regular expressions are needed, which is my point.
Look at your use case and data carefully and choose your solution
accordingly. If you generously use widespread regular expressions in many
cases you don't really need them, you'll likely not be happy with
performance. On the other hand, there are cases where regular expressions
are without a doubt the solution of choice.

Mike

date: Thu, 22 Dec 2016 14:16:42 -0500

from: dlclark@xxxxxxxxxxxxxxxx
subject: Re: SQL function equivalent to TestN?

"RPG400-L" <rpg400-l-bounces@xxxxxxxxxxxx> wrote on 12/22/2016 02:01:15
PM:

This technique will likely perform better than using regular expressions

to

test for a valid numeric:

values length( trim( translate( '123,456,789.00-', ' ',

'01234567890,.-', '

' ) ) )

Likely performed better, yes. but with much less accuracy. I'm
not belittling the idea, but the problem is that TRANSLATE can't tell the
difference between one decimal character and multiple, one sign character
or multiple, national delimiters, or embedded blanks vs. leading/trailing
blanks. Not all of the following are numeric but that method of using
TRANSLATE will treat them all as such.

'-123,456,789.00'
'123,456,789.00-'
'123.456.789,00-'
'.123 .456 .789'
'123-456-7890'

Sincerely,

Dave Clark