$window_start = 5;
$window_size = 5;
$text_body = "Please ignore all this text, this is the text wanted, pl
+ease none of this text";

if ($window_start > 0) {
        $text_body =~ /\S+/g foreach 1..$window_start;
        $start_index = pos($text_body); 
}  else {
        $start_index = 0;
}

$text_body =~ /\S+/g foreach 1..$window_size; 

$end_index = pos($text_body); 

$windowed_text = substr($text_body, $start_index, ($end_index - $start
+_index));

print $windowed_text;

#prints " this is the text wanted,"
[download]

See this? This is my head banging into the wall. Wall, meet head. Head, wall.

Here's what the problem seems to be: when I tested it, it was with a short string and a window large enough that the string ended well before the end of the window. IE, a 20 word string with a 100 word window.

Argh. I'm not sure where to go from here. Perhaps coffee is a good idea.

The whiles in my update should bail out when they run out of matches.

You may test on some input but it seems to take equal or less than window just fine

It seems you want a window x words big?

@words=$text_body =~ /(\S+)/g;
#compute $start and $end
$newtext=join(" ",@words[$start..$end]);
[download]

Update:Without copying. Note: I could not get the foreach deal to work at all so had to use while with a counter.

# $start and $window

$i++ while($i<$start && /(\S+\s+)/g);

$i=0;
while(/\G(\S+\s+)/g && $i<$window_length)
{
    $newtext .= "$1";
    $i++;
}
[download]

Note also: Two types of while. Probably should pick one you like the best and standardize on that.

This skips the $start number of words and the assigns $window_length words to $newtext preserving whitespace

Another Update: $start of 0 was not working but reversing the tests in the first while fixes that.

Ah, left out a relevant bit of info: I need to preserve whitespace. So what I *really* need is the index of the end of the $window_start'th word, and the index of the ($window_start + $window_size)'th word.

And these are potentially 1 meg strings, so I'd like to avoid copies if possible.

Doesn't seem to work: Given $window_start of 0, $window_size of 100, and $text_body of 'Testing news submission.', it gives 'news'.

It also seems to make....wait. Ok, I know where Testing is going and can fix that. It's not working on teh last word because there's no whitespace after it. Would changing \s+ to \s* work?

I think it would be safest (and safe to assume) that a 1 Meg string will be stored in a file and terminated with some sort of CR and or LF and in fact the pattern match I have would require that to be the case. The pattern would get a lot more complex if you could end a string with a \S type of entity.

However you have noticed a problem with the original code with a start of 0. I will update that bit.

Would something like this be simpler?

$window_size--;
my( $windowed_text )= $text_body =~
    /^\s*(?:\S+\s+){$window_start}(\S+(?:(\s+\S+){0,$window_size})/;
[download]

Note that my technique can also give you the same index information via @- and @+ if you have Perl v5.6 or higher.

Very nice solution, tye!!

There's a small typo in the regex (a surplus parenthesis), so here goes the code again, corrected

$window_size--;
my( $windowed_text )= $text_body =~
    /^\s*(?:\S+\s+){$window_start}(\S+(?:\s+\S+){0,$window_size})/;
[download]

What I wanted to add is that Perl handles the boundary cases very nicely, so no extra handling required for $window_start = 0 or $window_size = 0. This means

$_ = q/01234/;
/^..{0}/;    # matches '0'
/^..{0,0}/;  # matches '0'
/^..{0,-1}/; # doesn't match at all
[download]

which is exactly what we need for the code to work fine.

-- Hofmator

Thanks. Note that $window_size of 0 will probably behave the same as $window_size of 1, though.

Update: Sorry, wrong. I misunderstood your examples. /..{0,-1}/ matches any two characters followed by the literal string "{0,-1}". This could still be a problem for certain input values, but such seem pretty unlikely. So you might want some special code for the boundary case, depending on how varied your inputs might be.

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

What if $window_start > 32767 ?

It would seem with an average word length of say 5 that you could get more than 174,000 words per Megabyte of string input.