# Usage: $text = links($text);
sub links {
require URI::Find;
URI::Find->import();
local($text) = @_;
## Remove existing link code
# Patch # 1, </a> wasn't being caught so I convert the to a whitespace
$text =~ s#</a> #</a> #g;
while ($text =~ m#(<a[^>]* href=(["']).*?\2[^>]*?>(.*?)</a>)#gi) {
$pattern = $1;
$repl = $3;
$pattern = quotemeta($pattern);
if ($repl =~ /http/i || $pattern =~ /rel=["']nofollow/i) {
$text =~ s#$pattern#$repl#gsi;
}
}
## Convert www to http://www
# Patch # 2, trying to auto-correct www.example.com to http://www.example.com
# one problem, there's one site that uses a link like
# http://www.example.com/foo/www.example.com/bar.pdf
# which becomes
# http://www.example.com/foo/http://www.example.com/bar.pdf
# so the link is broken
# lookbehind is fixed-length, so I have to use 2 lines to get both http and https
$text =~ s#(?<!http://)www\.([a-z])#http://www\.$1#gi;
$text =~ s#(?<!https://)www\.([a-z])#https://www\.$1#gi;
# maybe use a \b word boundary?
# s#\b(?<!https://)www\.([a-z])#https://www\.$1#gi;
## Create Links
$finder = URI::Find -> new(
sub {
($uri, $orig_uri) = @_;
$uri =~ s/ *$//;
$orig_uri =~ s/ *$//;
# Patch # 3, ignore images from my site so I don't end up with
# <img src="<a href='blah' target='_new'>blah</a>">
if ($orig_uri =~ /$home(.*?)\.(jpg|jpeg|png|gif|bmp)$/i &&
$orig_uri !~ /cache/i) { return $orig_uri; }
else {
# Remove utm_, then trailing ? or &
$uri =~ s#utm_\w+=[^&]+(&(amp;)*)*##gi;
$uri =~ s#[?&]$##;
$orig_uri =~ s#utm_\w+=[^&]+(&(amp;)*)*##gi;
$orig_uri =~ s#[?&]$##;
# turn long links into something like https://www.example.com/.../bar/
if (length($orig_uri) > 40) {
$orig_uri = substr($orig_uri, 0, 27) . '...' . substr($orig_uri, -10);
}
return "<a href='$uri' target='_blank'>$orig_uri</a>";
}
}
);
$finder -> find(\$text);
# Patch # 4, fix nested links
$text =~ s#<a([^>]*?) title=(["'])*.*?\2 ([^>]*?)>#<a$1 $3>#gi;
$text =~ s#<a([^>]*?) href=(["']*)<a href=["']*(.*?)["']* target=["']*_new["']*>.*?</a>["']*([^>]*?)>(.*?)</a>
#<a href=$2$3$2$4$1>$5</a>#gi;
# Patch # 5, fix images
# maybe modify # 3 and remove $home(.*?) ? I'm not sure why they're separated
$text =~ s#<img([^>]*)src=["']*\s*<a[^>]*[ ]href=["']*\s*([^'"]*)["']*[^>]*>([^<]*)</a>['"]*([^>]*)>
#<img src="$2"$1$4>#gi;
# Patch # 6, I occasionally see https://http// and I'm not sure why
$text =~ s#https://http:*//#https://#gi;
return $text;
}
lookbehind is fixed-length, so I have to use 2 lines to get both http and httpsIs this really necessary? How often in actual practice will ://www\. be preceded by anything other than http or https? In any case can't you use pipes? The RegEx engine I spend most time with will happily accept (?<=http://|https://). Sometimes it's simpler just to capture: a simple (https?://) will save you several bytes.
I'm reminded of a bit of boilerplate that we used to have to drag out fairly often in the Apache subforum, about how you should start by explaining in English what you're trying to do, rather than jump directly into the code. What's the input you are working with, and what's the desired output?
Is this really necessary? How often in actual practice will ://www\. be preceded by anything other than http or https? In any case can't you use pipes? The RegEx engine I spend most time with will happily accept (?<=http://|https://). Sometimes it's simpler just to capture: a simple (https?://) will save you several bytes.
(?<!https*://)
I don't remember exactly what happened, it's just in my notes that "it didn't work".Sometimes we would then read you the riot act for simply saying “It doesn’t work” without further explanation ;) but here I am fully prepared to believe it, because lookbehinds have to be of fixed length. I'm not sure if this is universally true of all RegEx engines, but it applies to all that I've used. (We won't talk about javascript, which doesn't support lookbehinds at all. Ugh.)
what does the ? in (https?://) do?It makes the s optional. It's what you would have said in the lookbehind if you were allowed to do so, darn it. You may have been led astray by the ?: sequence, since (?:blahblah) would have special meaning, but in all other environments a colon is just a colon--here, the literal colon in the full URL.
