Re: Regular expression (regcomp()) ccsid issue

[Thomas Raddatz <thomas.raddatz@xxxxxx>]

Hi Scott,

That would pretty well explain why "\s" does not work. So far I could not find any documentation of regcomp() or regexec() explaining the supported options. The only hint is "The functions regcomp(), regerror(), regexec(), and regfree() use regular expressions in a similar way to the UNIX awk, ed, grep, and egrep commands.".

"egrep" supports "\s", at least on Linux boxes I have access to.

Regards,

Thomas.

-----Ursprüngliche Nachricht-----
Von: RPG400-L [mailto:rpg400-l-bounces@xxxxxxxxxxxxxxxxxx] Im Auftrag von Scott Klement
Gesendet: Sonntag, 3. November 2019 19:43
An: rpg400-l@xxxxxxxxxxxxxxxxxx
Betreff: Re: Regular expression (regcomp()) ccsid issue

Thomas,

Classes such as \s are not a standard part of regular expressions.  Can 
you point me to documentation that states that regcomp() supports them?  
I am very skeptical that regcomp() supports nonstandard regular 
expression classes like these.

I believe these were part of the Perl extensions to regular 
expressions.  Java implemented them as well.   But they are not 
available elsewhere except where the perl extensions are available.

-SK


On 11/3/19 10:10 AM, Tools/400 wrote:
> Hi,
>
> Hopefully somebody is smarter than me. The iSphere source member search
> (using regular expressions) does not work when using character classes,
> such as "\s".
>
> Whatever I try, I cannot make it work.
>
> I tried to convert the pattern and the value to ccsid 37 without success.
>
> I tried retrieving the current ccsid with nl_langinfo(CODESET) and
> convert the pattern and value to that ccsid without success.
>
> Last but not least setting the locale with setLocale(LC_ALL:
> LC_C_GERMANY) did not help.
>
> Actually I wanted to change the source file search to use SQL function
> REGEXP_COUNT. But that one is sooooo sloooow, that nobody wants to use
> it. But at least it works.
>
> REGEXP_COUNT takes 1644298 milliseconds when regexec() does the same job
> in 8118 milliseconds. That is 200 times faster than SQL!!!
>
> It was incredible if somebody could help me wit that. A working example
> running on jobs with ccsid 37 and 273 would be great.
>
> Regards,
>
> Thomas.