Esteemed Monks,

I am trying to translate the following Node.js fragment to Perl (it takes a URL and sort of chdir to it, clever enough to remove endpoint if a file and query):

const url = require('url');
const x = 'http:/a.com/a/b/c/d/e.json?x=1&y=2';
const y = '../../../';
console.log(url.resolve(x, y));
[download]

result is : http:/a.com/a/

Now, if x='http:/a.com/a/b/c/d/e.json/?x=1&y=2'; (i.e. the path points to a "directory" rather than to a "file") the result is http:/a.com/a/b/ (as expected IMO)

I tried to do this with URI but failed:

use URI;
my $x = 'http:/a.com/a/b/c/d/e.json?x=1&y=2';
my $y = '../../../';
my $u = URI->new($x);
my $p = $u->path().'/'.$y;
$u->path($p);
print $u;
<c>
<p>With this result: <c>http:/a.com/a/b/c/d/e.json/../../../?x=1&y=2
[download]

The Node.js version seems to be able to understand that the last segment before the query is a "file" and removes it. Then it does the merging of the two paths including successful relative operations (../). What is the safest way to achieve that behaviour in Perl? The last resort would be to shell-out to Node.js ...

bw, bliako