Regex Substitution Evaluations

mobiusinversion has asked for the wisdom of the Perl Monks concerning the following question:

Dearest Monks,

Please consider the following code:

use strict;

my $my_name = "august ferdinand mobius";

$my_name =~ s/\b(\w)/\U$1/g;

print $my_name;

#prints August Ferdinand Mobius
[download]

I am trying to create a subroutine that consumes a string, a pattern, a replacement and a modifier (perhaps limited to one or more of 'igsx') and returns the result of a substitution:

sub replace {
    my($text, $pattern, $replacement, $modifiers) = @_;
    (my $result = $text) =~ s/$pattern/$replacement/$modifiers;
    return $result;
}
[download]

I would like to call such a subroutine as follows:

$my_name = replace("august ferdinand mobius","\b(\w)","\U$1","igs");
[download]

Of course, the above doesn't work. But I have tried many other ways, e.g. using eval, s///ge, s///gee, etc.

I have attempted to do my homework using the Owl, Camel and Ram books, plus the perldoc, but I am totally stumped. Any help would be greatly appreciated.

Update: 13:06 PST
This appears to work:

use strict;

my $x = "august ferdinand mobius";
my $y = replace($x, '\b(\w)', '\U$1', 'igs');
print $y;

sub replace {
    my($text, $pattern, $replacement, $modifiers) = @_;
    my $result = '';
    eval '($result = $text)'."=~ s/$pattern/$replacement/$modifiers";
    return $result;
}
[download]

It does however, to my mind, open security problems due to the eval'ing - which I hope is unnecessary with a more elegant solution. Besides security, will this fail on certain edge cases? (besides non-escaped members of the dirty dozen, which I'm handling now)

Comment on Regex Substitution Evaluations Select or Download Code

Replies are listed 'Best First'.
Re: Regex Substitution Evaluations by pc88mxer (Vicar) on Jul 05, 2008 at 20:27 UTC
Evaling strings is fraught with numerous pitfalls, and I would avoid that approach. Most modifiers can also be expressed directly in the regular expression. E.g.: `m/foo/i; # -> m/(?i)foo/; m/bar/m; # -> m/(?m)bar/; ...` [download] See `perldoc perlre` for more details. So either add them yourself, or require that the caller do so.	[reply] [d/l] [select]
Re^2: Regex Substitution Evaluations by mobiusinversion (Beadle) on Jul 05, 2008 at 20:34 UTC
Please refer to my original post. As you can see, I would absolutely like to avoid using eval, and I have consulted perldoc's perlre perlop perlretut and perlreref entries several times. How might you suggest going about substitutions?	[reply]
Re^3: Regex Substitution Evaluations by Your Mother (Archbishop) on Jul 05, 2008 at 20:51 UTC
I think pc88mxer answered your question for the most part. `$my_name = replace("august ferdinand mobius","\b(\w)","\U$1","igs"); # Becomes something more like... $my_name = replace("august ferdinand mobius", qr/(?igs)\b(\w)/, '\U$1' +);` [download] Update: You should consider that your solution will not work for a variety of names, like mccleod, vomdorp, decaro, etc. For proper names, a dictionary (with the right library under it, like proper names) is a better way to go, and there is no perfect solution since the same spellings are sometimes capitalized differently. :(	[reply] [d/l]
Re^4: Regex Substitution Evaluations by oko1 (Deacon) on Jul 05, 2008 at 22:31 UTC
Re^4: Regex Substitution Evaluations by mobiusinversion (Beadle) on Jul 07, 2008 at 18:53 UTC
Re: Regex Substitution Evaluations by ikegami (Patriarch) on Jul 05, 2008 at 23:41 UTC
If the capture works, I'd use: `sub replace { my($text, $pattern, $replacement, $global) = @_; my $result = $text; if ($global) { $result =~ s/$pattern/$replacement/eg; } else { $result =~ s/$pattern/$replacement/e; } return $result; } my $y = replace($x, qr/\b(\w)/is, '"\U$1"', 1);` [download] Keep in mind the 3rd argument is arbitrary Perl code instead of a string literal. It used to be able to include arbitrary Perl code, so nothing's new on the (lack of) security front. If you really do want the options and/or regexp to be a string, I'd use my previously posted solution.	[reply] [d/l]