Can the nested loops be avoided? That's a good question.

The problem is that I have to map between the words I'm comparing and the original "un-modified" text.

Anyway, here's an example that runs.

http://hank.org/modules/PhraseTest.pm

(I think PerlMonk's wrapping code is too agressive in code blocks)

> dprofpp -u -p PhraseTest.pm

Total Elapsed Time = 0.506468 Seconds
         User Time = 0.508234 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c  Name
 61.0   0.310  0.430     30   0.0103 0.0143  PhraseHighlight::highligh
+t_text
 15.7   0.080  0.080     30   0.0027 0.0027  PhraseHighlight::build_hi
+ghlighted
                                             _text
 5.90   0.030  0.030     30   0.0010 0.0010  PhraseHighlight::substitu
+e_highlig
                                             ht
 3.94   0.020  0.020     15   0.0013 0.0013  Text::Wrap::wrap
 1.97   0.010  0.010     30   0.0003 0.0003  PhraseHighlight::join_wor
+ds
 1.97   0.010  0.010      1   0.0100 0.0100  warnings::BEGIN
 1.97   0.010  0.020      1   0.0100 0.0200  vars::BEGIN
 1.97   0.010  0.010      2   0.0050 0.0050  PhraseHighlight::BEGIN
 0.00   0.000 -0.000      2   0.0000      -  vars::import
 0.00   0.000 -0.000      3   0.0000      -  Text::Tabs::BEGIN
 0.00   0.000 -0.000      2   0.0000      -  result::BEGIN
 0.00   0.000 -0.000      4   0.0000      -  constant::BEGIN
 0.00   0.000 -0.000      1   0.0000      -  warnings::register::impor
+t
 0.00   0.000 -0.000      2   0.0000      -  warnings::register::mkMas
+k
 0.00   0.000 -0.000      1   0.0000      -  strict::unimport
[download]

Thanks,