Beefy Boxes and Bandwidth Generously Provided by pair Networks
Keep It Simple, Stupid
 
PerlMonks  

Re: Regex for outside of brackets

by mr_mischief (Monsignor)
on Jul 13, 2018 at 22:03 UTC ( #1218466=note: print w/replies, xml ) Need Help??


in reply to Regex for outside of brackets

Are you sure you want a match, and that you can only use a single regex? I have bad news... a regular language has no context. However, a regular expression and another tool or handful of tools can easily get you there. Take, for example, the substitution operator, a counter with a loop and some more regexes, or a regex match and a split on the match... Of course, feel free to use Text::Balanced as atcroft suggests or use some other toolset built for the level of the problem you're trying to solve. Regexes will only solve a subproblem of your problem.

Here's the data file for the following examples.

THIS IS OUTSIDE (THIS IS INSIDE) (inside) outside before (within) after before (within) between (within again) after b ((nested)) a before (within (nested)) after This one hangs (with an unmatched open This one has () an empty pair This opens (with one)) and double closes this is the last (really) one

Now here's the first example, using the substitution operator.:

#!perl use strict; use warnings; use 5.12.0; my $cleanup = 1; while ( <> ) { chomp; s/\(+.*?\)+//g; y/ / /s, s/^\s|\s$// if $cleanup; say; }

The above code produces the following output by substituting 0 characters in place of any pair of parentheses with anything between them. As written, it eliminates matched pairs and their contents but will also eliminate an extra closing parenthesis and will include in the output an opened but not closed parenthetical.:

THIS IS OUTSIDE outside before after before between after b a before after This one hangs (with an unmatched open This one has an empty pair This opens and double closes this is the last one

Or if prefer to preserve whitespace as it was, set $cleanup to 0

This next example produces mostly the same output as the ˘cleanup = 0 version of the above. It does so by counting nesting level of the parentheses after splitting the string into an array of characters. It then appends to the output string if the nesting level is 0 (outside of any pairs of parentheses). This one will produce its last output before a hanging opened and unclosed pair. It will, as written, also not include in the output negative nesting levels (text trailing an extra close unmatched by an open).

#!perl use strict; use warnings; use 5.12.0; while ( <> ) { chomp; my $str = ''; my @parts = split //; my $inside = 0; for ( @parts ) { /\(/ && $inside++ && next; /\)/ && $inside-- && next; $str .= $_ unless $inside; } say $str; }
THIS IS OUTSIDE outside before after before between after b a before after This one hangs This one has an empty pair This opens this is the last one

Or if you want to feed the match from one regex match into a split on that match...

#!perl use strict; use warnings; use 5.12.0; while ( my $str = <> ) { chomp $str; my $extract = join '|', map { "\Q$_\E" } ( $str =~ m/(\(+.*?\)+)/g + ); say join '', split /$extract/, $str; }

The above works because we know what we want to eliminate, which is a good use for split. In this particular case, we don't have a fixed regex against which to split, but we know how to match what we don't want. This solution captures that unwanted part, quotes it with \Q and \E, joins any multiples with the regex alternation (pipe, or '|'), then uses split and join to leave what's left of the string as a single string. This as written will only eliminate matched pairs and their contents. This is basically emulating the substitution operator. One's intuition may be that since it's a more detailed treatment it'll be faster. However, we're more Perl here, and the substitution operator is highly optimized. I don't know without benchmarking by how much, but I'm betting the example with the s/// is faster.

The second example above is fairly easy to tweak to give the sort of error messages you might expect out of a lexer, since it is kind of a degenerative case of one.

#!perl use strict; use warnings; use 5.12.0; my $inside = 0; while ( <> ) { chomp; my $str = ''; my @parts = split //; $inside = 0; for ( @parts ) { 0 > $inside && $inside++ && warn "WARNING: extra close on line + $.\n"; /\(/ && $inside++ && next; /\)/ && $inside-- && next; $str .= $_ unless $inside > 0; } warn "WARNING: Unclosed parenthetical on line $.\n" if $inside; say $str; }
THIS IS OUTSIDE outside before after before between after b a before after WARNING: Unclosed parenthetical on line 7 This one hangs This one has an empty pair WARNING: extra close on line 9 This opens ) and double closes this is the last one

Replies are listed 'Best First'.
Re^2: Regex for outside of brackets
by choroba (Archbishop) on Jul 13, 2018 at 22:20 UTC
    You can get the same output as in the second example by slightly modifying the substitution in the first example and calling it in a loop:
    1 while s/\([^()]*\)//g;

    To get the warnings, just check for remaining ( or ).

    ($q=q:Sq=~/;[c](.)(.)/;chr(-||-|5+lengthSq)`"S|oS2"`map{chr |+ord }map{substrSq`S_+|`|}3E|-|`7**2-3:)=~y+S|`+$1,++print+eval$q,q,a,

      My goal wasn't particularly to force them all to the same output. I'm yet to be sure what output OP wants beyond the single case of input. It was to demonstrate that there are multiple ways to do things, none of which are a single regular expression match. Thanks for helping expand on that.

Re^2: Regex for outside of brackets
by mr_ron (Hermit) on Jul 16, 2018 at 00:31 UTC

    Are you sure you want a match, and that you can only use a single regex? I have bad news... a regular language has no context. ... Regexes will only solve a subproblem of your problem.

    from ISBN 9780596004927: Programming Perl (4th Ed), Pattern Matching - Chapter 5, p. 167

    If you're acquainted with regular expressions from some other venue, we should warn you that regular expressions are a bit different in Perl. First they aren't entirely "regular" in the theoretical sense of the word, which means they can do much more than the traditional regular expressions taught in Computer Science classes.
    Mentioned similarly in Wikipedia

    Brackets with Perl regex is a FAQ, Can I use Perl regular expressions to match balanced text?, and is also covered in Rosetta Code. The regex techniques used are relatively new(er) and somewhat advanced but other answers have discussed modules that hide the technique behind a simple interface.

    Correction based on feedback from mr_mischief below:

    I also studied "regular languages"/"regular sets" in computer science and, understand the confusion while understanding the intention to avoid unneeded extensions, worry that ignoring all extensions to the more formal concept of a regular expression will cause more confusion. The Wikipedia article mentions backreferences as an example of a very commonly used extension. I don't really see the difference covered in perlre , perlrequick or perlretut and would be interested in any suggestions on finding or adding the information for any of those documents.

    Ron

      I hand-waived over Perl regexes not being entirely regular. They are still not (without doing really convoluted things) a Turing complete language. Matching the complement of the text that's bracketed is still a step beyond matching bracketed text. Just because one can hammer a screw does not make a hammer a screwdriver nor a screw a nail.

      I remember a time when helping people with questions that this community would actually try to offer advice on good, clean ways to do things. I'm unsure of the value of arguing that because something is strictly possible that we should help do things the wrong way.

Re^2: Regex for outside of brackets
by theravadamonk (Scribe) on Jul 17, 2018 at 05:20 UTC

    Hmm, Many codes. things to learn. Thanks a LOT for your wonderful efforts.

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://1218466]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others imbibing at the Monastery: (3)
As of 2020-09-29 03:50 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?
    If at first I don’t succeed, I …










    Results (145 votes). Check out past polls.

    Notices?