> However, this does NOT work as I would expect:

Really? Well swear words having word boundaries is what I expect.

> It seems like the profanity patterns have start of word / end of word anchors built into the patterns,

well it seems so, why don't you just dump the regex to be sure?

Personally I wouldn't want words like Essex to be flagged. (Or Dickens or zaddick)

> don't work if the word is embedded inside another string? Is there any way to control this behavior? 

After browsing thru the code ...

http://cpansearch.perl.org/src/ABIGAIL/Regexp-Common-2016060801/lib/Regexp/Common/profanity.pm

I saw this

 
pattern name => [qw (profanity)], create => '(?:\b(?k:' . $profanity .
+ ')\b)', ;
[download]

So I doubt there is any possible flag to disable the hard coded \b meta character.

But if you really need this feature you could just copy the code into your own subclass and change the pattern to your needs.

> because of the Scunthorpe Problem.

I once ran in a similar phonetic problem after mentioning Kant in an English conversation :)

Cheers Rolf
_{(addicted to the Perl Programming Language and ☆☆☆☆ :)
Je suis Charlie!}

You might try to trim the boundary assertions off of the stringized Regexp object (sorry for all the wrap-around):

c:\@Work\Perl\monks>perl -wMstrict -le
"use Regexp::Common;
 ;;
 print qq{$RE{profanity}};
 print qq{A: match '$1'} if 'xxxpissxxx' =~ m{ ($RE{profanity}) }xms;
 ;;
 print '--------';
 (my $erp = $RE{profanity}) =~ s{ \A \Q(?:\b\E (.*) \Q\b)\E \z }{$1}xm
+s;
 print qq{'$erp'};
 ;;
 print qq{B: match '$1'} if 'xxxpissxxx' =~ m{ ($erp) }xms;
"
(?:\b(?:(?:piss(?:\ take|\-take|take|e(?:rs|[srd])|ing|y)?|quims?|shit
+(?:t(?:e(?:rs|[dr])|ing|y)|e(?
:rs|[sdry])|ing|[se])?|t(?:urds?|wats?)|wank(?:e(?:rs|[rd])|ing|s)?|a(
+?:rs(?:e(?:\ hole|\-hole|hole|
[sd])|ing|e)|ss(?:\ holes?|\-holes?|ed|holes?|ing))|b(?:ull(?:\ shit(?
+:t(?:e(?:rs|[dr])|ing)|s)?|\-s
hit(?:t(?:e(?:rs|[dr])|ing)|s)?|shit(?:t(?:e(?:rs|[dr])|ing)|s)?)|low(
+?:\ jobs?|\-jobs?|jobs?))|c(?:
ock(?:\ suck(?:ers?|ing)|\-suck(?:ers?|ing)|suck(?:ers?|ing))|rap(?:p(
+?:e(?:rs|[rd])|ing|y)|s)?|u(?:
nts?|m(?:ing|ming|s)))|dick(?:\ head|\-head|ed|head|ing|less|s)|f(?:uc
+k(?:ed|ing|s)?|art(?:e[rd]|ing
|[sy])?|eltch(?:e(?:rs|[rsd])|ing)?)|ha(?:rd[\-\ ]?on|lf(?:\ a[sr]|\-a
+[sr]|a[sr])sed)|m(?:other(?:\
fuck(?:ers?|ing)|\-fuck(?:ers?|ing)|fuck(?:ers?|ing))|uth(?:a(?:\ fuck
+(?:ers?|ing|[aaa])|\-fuck(?:er
s?|ing|[aaa])|fuck(?:ers?|ing|[aaa]))|er(?:\ fuck(?:ers?|ing)|\-fuck(?
+:ers?|ing)|fuck(?:ers?|ing)))|
erde?)))\b)
--------
'(?:(?:piss(?:\ take|\-take|take|e(?:rs|[srd])|ing|y)?|quims?|shit(?:t
+(?:e(?:rs|[dr])|ing|y)|e(?:rs|
[sdry])|ing|[se])?|t(?:urds?|wats?)|wank(?:e(?:rs|[rd])|ing|s)?|a(?:rs
+(?:e(?:\ hole|\-hole|hole|[sd]
)|ing|e)|ss(?:\ holes?|\-holes?|ed|holes?|ing))|b(?:ull(?:\ shit(?:t(?
+:e(?:rs|[dr])|ing)|s)?|\-shit(
?:t(?:e(?:rs|[dr])|ing)|s)?|shit(?:t(?:e(?:rs|[dr])|ing)|s)?)|low(?:\ 
+jobs?|\-jobs?|jobs?))|c(?:ock(
?:\ suck(?:ers?|ing)|\-suck(?:ers?|ing)|suck(?:ers?|ing))|rap(?:p(?:e(
+?:rs|[rd])|ing|y)|s)?|u(?:nts?
|m(?:ing|ming|s)))|dick(?:\ head|\-head|ed|head|ing|less|s)|f(?:uck(?:
+ed|ing|s)?|art(?:e[rd]|ing|[sy
])?|eltch(?:e(?:rs|[rsd])|ing)?)|ha(?:rd[\-\ ]?on|lf(?:\ a[sr]|\-a[sr]
+|a[sr])sed)|m(?:other(?:\ fuck
(?:ers?|ing)|\-fuck(?:ers?|ing)|fuck(?:ers?|ing))|uth(?:a(?:\ fuck(?:e
+rs?|ing|[aaa])|\-fuck(?:ers?|i
ng|[aaa])|fuck(?:ers?|ing|[aaa]))|er(?:\ fuck(?:ers?|ing)|\-fuck(?:ers
+?|ing)|fuck(?:ers?|ing)))|erde
?)))'
B: match 'piss'
[download]

Update: Of course, this gets you right back to the Scunthorpe Problem noted above by Paladin!

I followed your suggestion and tried this:

use strict;
use Regexp::Common;

(my $reg = $RE{profanity}) =~ s{\A \Q(?:\b\E (.*) \Q\b)\E \z}{$1}xms;

while ( my $word = <DATA> ) {
  chomp $word;
  if ( $word =~ m/$reg/ ) {
    print "Profanity detected: \"$word\"\n";
  } else {
    print "$word\n";
  }
}


__DATA__
aaaabbbbcccc
aaaashitcccc
aaaa1234cccc
ddddeeeeffff
[download]

This way it will find embedded "bad words" without the need for spaces around them, which is what I wanted. I realize the logic in requiring the word boundaries. But I think the fact that $RE{num}{int} finds embedded numbers made me assume that $RE{profanity} should work the same way, or else there might be a switch to toggle the behavior one way or the other.

The reason I need this is to generate temporary (one-use) passwords (like when someone requests a password reset on a website). The generated password should, ideally, be a jumble of random letters and/or numbers, but I don't want to accidentally send someone a password with an "obvious" obscenity embedded, so a simple filter like this is helpful.

Thanks!

You might consider adding a test to check if the expected alteration to the original regex was successful. The  \Q(?:\b\E and  \Q\b)\E parts of the substitution are rather fragile IMO and may break if the maintainer(s) of Regexp::Common ever change his/her/their notion of what a proper profane regex should look like.

c:\@Work\Perl\monks>perl -wMstrict -le
"use Regexp::Common;
 ;;
 (my $reg = $RE{profanity}) =~ s{\A \Q(?:\b\E (.*) \Q\b)\E \z}{$1}xms
   or die 'profanity anchor trim failed';
 ;;
 print qq{bad: '$1'} if 'Matsushita' =~ m{ ($reg) }xms;
"
bad: 'shit'
[download]

---

Give a man a fish:  <%-{-{-{-<

Shouldn't you be generating passwords that do not contain any words?