[ABE.pm] Re: perl regex Q

[Faber Fedor]

On 14/12/04 17:16 -0500, Ricardo SIGNES wrote:
> The (?= ... ) construct is called a look-ahead assertion.  

And that is the Perl magic I was looking for.

> > Also, is there a way to do a perl "edit in place" inside of a perl
> > program other than using backticks (`perl -iBAK -p -e 's/,,/,\\N,/g'
> > foo.txt`) or the system() call?
> 
> No.

I guess that's why (different regex) 

    perl -iBAK -p -e 's/ *NA */\\N/g' foo.txt

works from the command line but doing it from inside a perl program
generates the error:

    Missing braces on \N{} at -e line 1, within string
    Execution of -e aborted due to compilation errors.

It doesn't matter how many slashes I put in front of the N! :-)

> Consult "perldoc perlrun" and look under the section about -i;  it
> explains how -i would be implemented in Perl, more or less.

I'll go read it forthwith.

> > (No, reading the file in a line at a time is not an option.) 
> Why not?  It's what perl does when you use -i.

Seems a helluva lot faster than when I tried it in a while loop a while
back.

> Give us more real-world information about what you need to accomplish,
> and what the limitations are.

I need to replace "NA" with "\N" (no quotes around either). The file is
900K+ lines long  and 136 columns wide and will continue  growing
lengthwise.  Limitation: time. If I can do the "edit-in-place" from
inside my program, the whole process takes a minute or two.  Everything
thing else I've tried (except a while loop) takes 40 minutes or longer.




-- 
 
Regards,
 
Faber                     

Linux New Jersey: Open Source Solutions for New Jersey
http://www.linuxnj.com