in reply to Regex with multiple pattern omissions

I'm having a bit of trouble understanding the question. It sounds like your routine does what you want? Or not? If not then what else should it do? If you are asking for a "better" way to accomplish what you already have, I would say don't bother. What you have so far is reasonable. It appears to me that you have a clear algorithm that you understand.
  • Comment on Re: Regex with multiple pattern omissions

Replies are listed 'Best First'.
Re^2: Regex with multiple pattern omissions
by jhoop (Acolyte) on Jan 09, 2011 at 00:46 UTC
    Thanks. Sorry if I was unclear. What I have does several things that I want, but not everything. I would like to augment the match conditions to keep anything between "" together as one string, and to omit the control words regardless of case.. I am wondering, if I declare @omissions before the match statement, is it possible to ?! its contents in the match expression, having the m/.../i case-insensitivity apply to the contents of @omissions (eventually @omissions will be user-defined and might contain different things). Also, while checking for dupes after the fact is fine for small arrays (and I'm generally ok leaving it this way) i was wondering if there's a neat (more efficient) way to do it as each matched term is added, in case the input list is huge..
[reply]
      Thanks, this is a lot more clear now!

      1. One of the very cool things about Perl is that you can build regexes dynamically - this works great. So this can play into the eventual plan for @omissions.

      2. Using hash table like you have is a very Perl way to remove dupes. This will work fine even for bigger arrays.

      Need to noodle on the regex part of your question...

[reply]
Re^2: Regex with multiple pattern omissions
by jhoop (Acolyte) on Jan 09, 2011 at 00:59 UTC
    the immediate issue is that, in the current output given - oR, Or, and NOT should be omitted (in this case "and" is the only control-word from the input string that was correctly omitted) and also, "non" and "volatile" should remain together in the output
[reply]
      eep. i meant "non$volatile display" should remain together in the output
[reply]
        Is this headed in the right direction?
        Meaning does this code produce the output that you want? (given your single test case)...Program specifications are hard to write and I think the best way forward here is just refinement by example. 
        #!/usr/bin/perl -w
use strict;

my $input = '"non$volatile display" and ((timer oR count$3 Or display)
+ near5 hour).ccls. NOT (LCD).ab.';

my @omissions = qw(terms and or not with near same xor adj);
my $omit = join("|",@omissions);

$input =~ s/\..*?\.//g;
$input =~ s/$omit//ig;
my @searchterms = ($input =~ m/".+?"|[a-zA-Z][\w\$]+/gi);
print "@searchterms";

#prints:
#"non$volatile display" timer count$3 display hour LCD

        [download]
        update:
        -I think that you mean for these .xyz. terms to be deleted?
        -Above doesn't allow for terms in @omissions to be taken absolutely literally. I need to look in Larry's book for the syntax. But this does show a dynamic regex. Also probably need to take into account that omit words should be on boundaries (whole words - not words within words, the \b - look in Larry's book)
        -Running substitute operations can take some time as the string is modified after each one, but this may or may not matter time-wise.
        -The question right now: is this is "right" output? I mean for this given single input case?

        As a general approach, I try to break these complex things into multiple easier steps. Get the right output, then tweak it if performance is not adequate.

[reply]
[d/l]

In Section Seekers of Perl Wisdom