Re^5: Regex: Matching around a word(s)

Okay, this will amalgamate two matches closely adjacent matches:

#! perl -slw
use strict;

my $word = $ARGV[0] or die "No search term";

( my $text = do{ local $/; <DATA> } ) =~ tr[\n][]d;

$text =~ s[
    ( (?: \S+ \s+ ){1,3} )
    ( $word ) [[:punct:]]*
    (?= (
        (?:
            (?: \s+ \S+ ){0,6}
            \s+ ( $word ) [[:punct:]]*
            (?: \s+ \S+ ){1,3}
        )
        |
        (?: \s+ \S+ ){1,3}
    ) )
][
    my $extract = "$1<$2>$3";
    $extract =~ s[\s($word)][ <$1>]g;
    print $extract;
]gex;

__END__
Regular expressions have always been a weak spot for me, and I've got 
+a
 question that's got me stumped. Here's the problem I'm trying to solv
+e.
 I have somwhat large articles of text (returned from a search), what 
+I'd
 like to do is capture the word and X number of words before and after
+ it
 while tagging the matching word in the captured text. My inital thoug
+ht
 was to try something like this. The problem I have is that if there i
+s
 more than one term and they overlap, the nth term will not be annotat
+ed.
 So my next thought is lookahead/lookbehind, but they don't capture.
 Is there a way to do this with a single regex? Is a regex even the be
+st
 way to do this? Thanks, -Lee
[download]

A test

P:\test>517393 and
spot for me, <and> I've got a
capture the word <and> X number of words before <and> after it while
of words before <and> after it while
than one term <and> they overlap, the
[download]

As is, it won't try for a third or fourth, (and is currently repeating itself!), but it should be possible to do that using a independant subregex ((?{ $regex ))), I just haven't got it right yet. I'll have another go tomorrow when my eyes are open :)

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

Lingua non convalesco, consenesco et abolesco. -- Rule 1 has a caveat! -- Who broke the cabal?

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

Comment on Re^5: Regex: Matching around a word(s) Select or Download Code

Replies are listed 'Best First'.
Re^6: Regex: Matching around a word(s) by shotgunefx (Parson) on Dec 17, 2005 at 03:10 UTC
I appreciate all the effort. I think to get the best results, I'll have to scan the text. Here's my first attempt. Seems like there should be a less klunky way though. #!/usr/bin/perl use strict; use warnings; die "No search terms supplied!" unless @ARGV; my @words = @ARGV; my $text; { local $/ = undef; $text = <DATA>; } my $regex = join("\|",@words) ; my $expr = qr /($regex)/; my $get_length = 100; # Characters before and after the end of match t +o grab. my (@found_pos, @grab_pos); while ($text =~ /($expr)/sg) # Find matches and save the positions [pr +e,post] in @found_pos, { my $ipos = pos($text); my $prepos = $ipos - $get_length > 0 ? $ipos - $get_length : 0; my $postpos = $ipos + $get_length < length($text) ? $ipos + $get_len +gth : length($text) ; push @found_pos, [$prepos,$postpos]; } my $first = shift @found_pos; # Get the first "span" while (@found_pos){ my $n = shift @found_pos; # Get the next "span" if ($first->[1] > $n->[0]){ # Overlap? $first->[1] = $n->[1]; # compact push @grab_pos, $first unless @found_pos; }else{ push @grab_pos, [@$first],$n; # No, seperate, push them onto sta +ck $first = $n; } } foreach (@grab_pos){ my $chunk = substr($text,$_->[0],$_->[1]) ; $chunk=~s/\b($expr)\b/[$1]/igs; print ">>>",$chunk,"<<<\n"; } __DATA__ Regular expressions have always been a weak spot for me, and I've got +a question that's got me stumped. Here's the problem I'm trying to solv +e. I have somewhat large articles of text (returned from a search), what + I'd like to do is capture the word and X number of words before and after + it while tagging the matching word in the captured text. My inital thoug +ht was to try something like this. The problem I have is that if there i +s more than one term and they overlap, the nth term will not be annotat +ed. So my next thought is lookahead/lookbehind, but they don't capture. Is there a way to do this with a single regex? Is a regex even the be +st way to do this? Thanks, -Lee [download] -Lee perl digital dash (in progress)	[reply] [d/l]
Re^7: Regex: Matching around a word(s) by BrowserUk (Patriarch) on Dec 17, 2005 at 13:04 UTC
Having run your code, it's not entiely clear to me exactly what you want as output. If you want the entire body of the text with the keyword highlighted, that's easy. If you want from the first to last match plus some context on either end that's also easy, but in both cases you end up extracting most if not all of the article if there are two occurrances of the match far apart. Alternatively, if you want a shortish snippet, (5 or 10 words tops), for each match, but with any snippets of the target length that include two or more matches merged, then this seems to work for that limited set of cases I've tried: P:\test>517393-3 -N=1 got ...I've <<got>> a question that's <<got>> me... P:\test>517393-3 -N=3 got ...me, and I've <<got>> a question that's <<got>> me stumped. Here's... P:\test>517393-3 -N=5 got ...spot for me, and I've <<got>> a question that's <<got>> me stumped. Here's the problem... P:\test>517393-3 -N=3 and ...spot for me, <<and>> I've got a... ...capture the word <<and>> X number of words before <<and>> after it while... ...than one term <<and>> they overlap, the... P:\test>517393-3 -N=3 the ...me stumped. Here's <<the>> problem I'm trying... ...do is capture <<the>> word and X... ...it while tagging <<the>> matching word in <<the>> captured text. My... ...and they overlap, <<the>> nth term will... ...a regex even <<the>> best way to... P:\test>517393-3 -N=3 of ...somwhat large articles <<of>> text (returned from... ...and X number <<of>> words before and... P:\test>517393-3 -N=10 of ...problem I'm trying to solve. I have somwhat large articles <<of>> text (returned from a search), wh +at I'd like to do... ...like to do is capture the word and X number <<of>> words before and + after it while tagging the matching word... [download] Maybe that's useful, but if not it was an interesting exercise :) Read more... (2 kB) Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error. Lingua non convalesco, consenesco et abolesco. -- Rule 1 has a caveat! -- Who broke the cabal? "Science is about questioning the status quo. Questioning authority". In the absence of evidence, opinion is indistinguishable from prejudice.	[reply] [d/l] [select]
Re^8: Regex: Matching around a word(s) by shotgunefx (Parson) on Dec 18, 2005 at 00:34 UTC
Nice++ Again, much appreciated. Originally, I wasn't thinking it quite through. I was just thinking I could grab excerpts around matches and highlight them. Then seeing the results, I decided it needed to merge overlapping excerpts into one. That regex is damn close though. If that was the first thing you posted, I would have been done right there :) With the sample data we've been looking at, it's quite easy to get the whole text. The real page results are rather long and it's unlikely that you'd get the whole page and many stop words aren't indexed. Though I'll probably add a limit that either truncates the results if the excerpt is too big or something along that lines. I found it an interesting (and consuming) excercise myself. -Lee perl digital dash (in progress)	[reply]