#!/usr/bin/perl
use warnings;
use strict;

use URI;
use Test::More tests => 2;

my %expect = (
    '/blah1/blah2/blah3/1234567890;arg=AAA123BBB456CCC' => '1234567890
+',
    '/blah1/blah2/blah3/1234567890' => '1234567890');

for my $string (keys %expect) {
    my $uri = ('URI'->new($string)->path_segments)[-1];
    is $uri, $expect{$string};
}
[download]

thanks. I'm not familiar with URI, but looks like it's the way to go.

How about m{/(\w+)(?:$|;arg=)}? (WebPerl Regex Tester)

However, obviously there may be cases where the regex might not work. As you've only given two examples, that's not really enough to confirm. Please see Re: How to ask better questions using Test::More and sample data for how to write more test cases for regexen. Or, you might want to look into using a proper parser such as URI (Update: choroba just posted an example).

thanks guys!

ciaoperl:

Yet another way you could do it:

It might look a little magical at first, but it's actually pretty simple. Instead of using a regular expression, I first split the string on ";" and kept only the first bit ((split ";")[0]) to chop off the optional bits. Then I split that first bit on "/" and kept the last part ((split "/", ...)[-1]) to get the value you wanted. Finally, I assigned the result to $_ and used the -p option to print the value.

I got there essentially by starting with something like this:

my $url = "/bla/bla/123;yarg";
my @tmp = split ";", $url;    # break string at semicolons
@tmp = split "/", $tmp[0];    # break first chunk apart at "/"
print $tmp[-1];               # Print the value we want
[download]

Since you can extract a chunk of a list by subscripting it, I first contracted the above to:

my $url = "/bla/bla/123;yarg";

# Split on ";", keep only the first chunk, then split on "/"
my @tmp = split "/", (split ";", $url)[0];

# The last item holds the value we want
print $tmp[-1];
[download]

Then I wanted to get rid of the temporary array by keeping *only* the final result by again subscripting the resulting list:

my $url = "/bla/bla/123;yarg";

$url = (split "/", (split ";", $url)[0])[-1];

print $url;
[download]

Then to turn it into a one liner, I changed it to this:

perl -e '$a="/bla/bla/123;yarg"; print (split "/", (split ";", $a)[0])
+[-1]'
print (...)_ interpreted as function at -e line 1.
syntax error at -e line 1, near ")["
execution of -e aborted due to compilation errors.
[download]

Oops! The parser sees "print (" and thinks we're wrapping the print arguments in parenthesis, leading to a syntax error later. I could've fixed the problem by putting a unary "+" on the thing to print, as that tells perl that the parenthesis isn't part of print:

perl -e '$a="/bla/bla/123;yarg"; print +(split "/", (split ";", $a)[0]
+)[-1]'
123
[download]

But I went a different route: since you're using echo to inject the value into perl as the value $_, I instead elected to use the -p option which tells perl to print the value in $_ after the code executes, giving me the bit of code I suggested.

...roboticus

When your only tool is a hammer, all problems look like your thumb.

Sometimes Perl is not the right tool for the job. You are starting an entire instance of perl just for a regex match, when sed will do:

echo "$XXX" | sed -e 's/;arg=.*$//' -e 's!^.*/!!'
[download]

On my box, simply running perl -e 1 takes longer than the above sed command. For a shell script, these differences matter.

Perl oneliners are great at the prompt, but if you are putting them into a shell script, you should probably move the entire script to Perl.

you have a good point jcb. For simple one-liners sed will do. Thanks!

echo "$XXX" | perl -ne '/\/(\w+);/ and print $1'
[download]

Unfortunately, semicolon ";" is not guaranteed, so this won't work -- the part starting with (and including) ";" may or may not appear in the url.
Any other suggestions?

In that case:

echo "$XXX" | perl -ne '/\/(\w+)[^\/]*?$/ and print $1'
[download]