Given the two updates, LTjake certainly has the approach I would take, but I think the version I saw would end up scanning the entire input post, rather than quiting as soon as the output string is done. Here's the way the while loop was written when I first saw it (with my commentary added):

while ( my $token = $p->get_token ) {
    if ($token->is_text) {
        if (length($token->return_text) + $total <= 200) {
            $doc2 .= $token->return_text;
            $total += length($token->return_text);
        }
        else {
            for (split / /, $token->return_text) {
                if ($total + length($_) <= 200) {
                    $doc2 .= $_ . ' ';
                    $total += length($_) + 1;
                }
                else {
                    last;  ## THIS ONLY EXITS THE FOR LOOP
                }          ## So this block runs over the
            }              ## entire remainder of the post
            chop($doc2) if $doc2 =~ /\s$/;
        }
    }
    else {
        $doc2 .= $token->as_is;
    }
}
[download]

The solution would be to add the length test to the while loop condition, or else figure a way to avoid an inner for loop, so that "last" will really finish things off. And some other nit-picks:

• Rather than calling the same "return_text" method so many times, I'd rather use it once to set a local variable (and having just read the TokeParser::Simple docs, it seems that "return_text" is depricated anyway -- use "as_is" in all cases).
• The for loop uses a split on single space -- not what the OP intended -- but this leads to a question about the OP's stated goals: Would you want every input string of /\s{1,}/ to be represented (and counted) as a single space in the output string? (Myself, I would think "yes".)

So here's my version of LTjake's while loop (not tested):

while ( my $token = $p->get_token ) 
{ 
    my $tkntext = $token->as_is;
    $tkntext =~ s/\s+/ /g;  # normalize all whitespace

    if ($token->is_text) {
        if (length($tkntext) + $total <= 200) {
            $doc2 .= $tkntext;
            $total += length($tkntext);
        }
        else {
            my $maxlen = 200 - $total;
            $doc2 .= substr( $tkntext, 0, rindex( $tkntext, ' ', $maxl
+en );
            last;  # this finishes the while loop
        }
    }
    else {
        $doc2 .= " $tkntext ";
    }
}
[download]