FYI on Posting at Perlmonks in-a-nutshell; put your paragraphs inside <p></p> tags, and your code inside <code></code>. In your browser, select "view source" on this page and look at this post. It should show you (kind of...) the extra tags and their placement, which you can compare with what this post actually looks like. Also, you can preview your post as many times as you like until it looks right

As you said yourself, the code you've written may be stupidly over-complicated... but I'm assuming you're posting here not because of the code itself, but because the code produces an apparent bug with the Perl interpreter...

I found the debug switch use re qw{eval Debug ALL};, gives the extra output below:

Compiling REx "(?(?{$_ =~ |$regex (\w+ \w+)(?=\1)|})(?{s|$1||})|(?{s|$
+1 \w+"...
Starting first pass (sizing)
 >(?(?{$_ =~... |   1|  reg
                |    |    brnc
                |    |      piec
                |    |        atom
 >?(?{$_ =~ ... |    |          reg
 >?{$_ =~ |$... |   2|            reg
 >(?{s|$1||}... |   6|            brnc
                |    |              piec
                |    |                atom
 >?{s|$1||})... |    |                  reg
 >(?{s|$1 \w... |  12|            brnc
                |    |              piec
                |    |                atom
 >?{s|$1 \w+... |    |                  reg
Required size 16 nodes
Starting second pass (creation)
 >(?(?{$_ =~... |   1|  reg
                |    |    brnc
                |    |      piec
                |    |        atom
 >?(?{$_ =~ ... |    |          reg
 >?{$_ =~ |$... |   2|            reg
panic: top_env

C:\Temp\panic>
[download]

Many here are reluctant to follow off-site links. Worse, such off-site postings are apt to be ephemeral... and don't make it into the Monastery db for future reference. Please see Markup in the Monastery, Perl Monks Approved HTML tags and What shortcuts can I use for linking to other information?.

Having overcome my own reluctance, here's the pastebin content as you have it formatted:

bash-3.1$ weather --weather Newcastle | ./regex.pl
panic: top_env

======================================================================
+=======
The output from (weather --weather Newcastle) the regex would be proce
+ssing:
======================================================================
+=======

bash-3.1$ weather --weather Newcastle
   Newcastle Light Rain Late Light Rain Late. Morning Clouds. Warm.

======================================================================
+=======
                ( regex.pl )
======================================================================
+=======

#!/usr/bin/perl -w
use strict;
my $regex = qr/\s* \w+/;
while(<>)
{
        use re 'eval';
        /(?(?{$_ =~ |$regex (\w+ \w+)(?=\1)|})(?{s|$1||})|(?{s|$1 \w+|
+|}))/;
}

======================================================================
+=========
                Conclusion
======================================================================
+=========

From what i can gather from lurking google, code sub-patterns are very
+ experimental
and bug prone, ofcourse i could be wrong, seeing as i've only recently
+ adopted
Perl in the last month once finishing my final year at school.

From lurking google once more, i'm also gathering that the error messa
+ge i'm 
receiving is a bug within perl itself, and should be reported.

Bibliography:
http://www.perlmonks.org/index.pl/?node_id=368099
http://prlmnks.org/html/383052.html
[download]

It would be better, of course, to enclose (only) the code (by sections) in <c>...</c> tags and the narrative in <p>...</p> tags.

Update: re desemondo's exploration,changing the use re... to use re qw(eval debug); and then testing via checking ( perl -c 813676.pl ) produced this for me (w2k, perl 5.8.8):

Compiling REx `(?(?{$data =~ |$regex (\w+ \w+)(?=\1)|})(?{s|$1||})|(?{
+s|$1 \w+||}))'
size 15 Got 124 bytes for offset annotations.
panic: top_env
[download]

urthwrm:

/(?(?{$_ =~ |$regex (\w+ \w+)(?=\1)|})(?{s|$1||})|(?{s|$1 \w+||}))/

perlre on (?{ }):

Because Perl's regex engine is currently not re-entrant, interpolated code may not invoke the regex engine either directly with m// or s///), or indirectly with functions such as split.

Additionally, changing the string against which you are matching when you're still matching against it is just asking for trouble.

So, since the 'code sub-pattern' won't work ( see ikegami's++ above ) in the way you've attempted (and ++ for taking the challenging route), you may wish to reconsider your approach to the climb.

If so, the following -- hidden in the readmore so as not to spoil the challenge you still face -- will work. Note, however, that I have assumed that the list of data returned by

bash-3.1$ weather --weather Newcastle
009:    Newcastle Light Rain Late Light Rain Late. Morning Clouds. War
+m.
[download]

represents a specific label ("Newcastle) followed by a set of possible alternatives

So, without further ado:

Sorry for the delay, I've been busy with university. I appreciate all the reply's, i have learned a lot. Anyway, i ended up taking the easy way out just to get things working. Here's my code for who ever might be interested!

 #!/usr/bin/perl -w

use strict;

while(<>)
{
        if(m/\s*\ \w+ 
        \ (\w+|
        \w+\ \w+|
        \w+\ \w+\ \w+) 
        \ (?=\1)/x)
        {
                $_ =~ s/$1//;
                print;
        }
}
[download]