/(?$modifiers)$pattern/
[download]

Or /(?$modifiers:$pattern)/ where $modifiers can be "i-sm", for example, to turn on "i" and turn off "s" and "m". The first form (without a ":" affects the rest of the regex. The second form only affects the stuff between the parens.

See "perldoc perlre" for more info. See also qr in "perldoc perlop".

        - tye (but my friends call me "Tye")

I was looking for a similar issue. My problem wasn't with a match, but a substitution. This node is fairly old, however very relevant to my problem and I'm adding my comments in case someone needs a similar answer.

<bold>Project details:</bold>
I'm building a "interactive shell" to make our life easier. (DBA Group) In one shell I plan to tie Oracle, Sybase (DBI) and Sybase Replication together. Replication is also managed via DBI, but I provide a customized syntax for manipulating it. (They don't provide bulk loading functionality for heterogeneous environments. Thats done via other Perl scripts/modules) ;) I wanted to provide the user with a way of manipulating the input "buffer" via regex.

The only modifiers I allow are i,s,m & g for simplicity. ('e' & 'm' would be more difficult to "grammerize" than I'm willing to spend time on) While i,s & m are fine using the syntax (?$mods), I don't believe the g is. Well, no big deal.. My solution:

if ($mods=~s/g//g) {
  $buffer=~s/(?$mods)$expr1/$expr2/g;
} else {
  $buffer=~s/(?$mods)$expr1/$expr2/;
}
[download]

Hope this helps someone else, like myself! Great thread, thanks to all!

Regards,
SMF 8)

You can also do it using an eval which is perhaps more obvious than no_slogan's method but not as elegant.

$_ = 'foo(';

$regex = 'FOO(';
$modifiers = 'i';

# a bit of error checking never goes astray
die "Illegal modifier!" unless $modifiers =~ m/^[ismxo]\z/;

# escape everything except the essential regex char and \w \s
# this is a security measure as discussed below
$regex =~ s/([^\w\s\\|()[\]{}^\$*+?.])/\\$1/g;

# generate the regex with error checking
$regex = eval "qr/$regex/$modifiers";
die "Eval failed $@\n" if $@;

# test it
print "match" if /$regex/;
[download]

There is a significant security issue eval-ing user supplied data, or in fact using raw user data in many places in a script. The problem is scary. If $regex contains a string like this '/; `some bad command`; $hacked =1' then when this interpolates we will eval

$regex = eval "qr//; `some bad command`; $hacked =1/$modifiers"

This may execute some bad command on your system with the same permissions as the script. Both the solutions presented suffer from this problem. To solve it (at least partially) you need to escape all the non regex metachars.

$regex =~ s/([^\w\s\\|()[\]{}^\$*+?.])/\\$1/g;
[download]

This will leave \w \s and the regex chars unescaped but throw in escapes for everything else including semicolons and backtics. It can probably still be hacked - did someone say merlyn - but the escapes should throw a spanner in the works for the casual hacker.

cheers

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print

$modifiers =~ tr/cgimosx//cd;  # remove illegal mods
$regex = eval qq{ qr/\$regex/$modifiers };
[download]