URLs are the same everywhere you go, so you don't need the localization magic of File::Spec to figure out how to put them together. Just use the forward slash (/) to join the path elements. It's so trivial I guess URI doesn't see fit to wrap it in a method name. :)

So, in your example, if $uri->path ends in a slash, just concatenate the foo.html. If the $uri->path doesn't end in a slash, make it end in a slash, then add the foo.html. That's just what File::Spec::Unix->catfile is doing, although you're not really supposed to be peeking down there. :) If that's all your using that module for, it's probably not worth it to even compile it. :)

of course, but my goal is to avoid the "if the $uri->path doesn't end in a slash do something different" part, if possible.

$uri->path( $uri->path =~ m{/$} ? $uri->path . 'index.html' : $uri->pa
+th . '/index.html')
[download]

and similar seems more clumsy.

what i actually end up with is typically more like $base .= '/' unless $base =~ m{/$};, but i'm looking for something smarter, because i hate repeating these things and the problem is so easily solved (use File::Spec) with other file paths outside of URI context.

as for localization, that's why i used File::Spec::Unix explicitly, because you are correct, i don't need localization magic.

Slightly less clumsy?

$uri->path( $uri->path
           .( $uri->path =~ m{/$} ? '' : '/' )
           .'index.html'
          );
[download]

$_="http://bla.com";
s#([^/])$#$1/#; 
$_ .= "foo.html";
print;
# prints http://bla.com/foo.html


$_="http://bla.com/";
s#([^/])$#$1/#; 
$_ .= "foo.html";
print;
# prints http://bla.com/foo.html
[download]

Update: Sorry, misread the OP. Anyways, it has a hint about it:

If you simply want to construct file URI objects from URI strings, use
+ the normal URI constructor. If you want to construct file URI object
+s from the actual file names used by various systems, then use one of
+ the following URI::file constructors:
...
[download]

that is for file://-schemed URIs, not http, and it doesn't do concatenation, so i have trouble seeing how it could be of use here, unless i am missing something.

#!/usr/bin/perl  -- 

use strict;
use warnings;

use URI;

my $base = 'http://foobert.org/dir/odir';
my $file  = 'foo.html';
my $uri = URI->new($base);

local $\=$/;
print $uri->path_segments( $uri->path_segments, $file );
print $uri;
print $uri->host('I-RTFM.org');
print $uri;

__END__
dirodir
http://foobert.org/dir/odir/foo.html
foobert.org
http://I-RTFM.org/dir/odir/foo.html
[download]

$uri->path_segments is interesting and useful, but it doesn't solve the "behave the same whether $base ends with a slash or not" problem.

change the code to my $base = 'http://foobert.org/dir/odir/'; and you get http://foobert.org/dir/odir//foo.html, with two slashes!

however, this will do it :)

print $uri->path_segments( grep { $_ } $uri->path_segments, $file );
[download]

uses grep to filter out the empty strings returned by path_segments.

updated: fixed typo