Don't omit "/" if it has QUERY_STRING.

[hiratara]

RFC 1738 says, `If neither <path> nor <searchpart> is present, the "/" may also be omitted.'.

http://www.rfc-editor.org/rfc/rfc1738.txt

So we can't omit "/" if the URL has '?' and QUERY_STRINGs.

I thought that the following patch is better than this pull-req, but it makes t/old-base.t broken.

diff --git a/URI/http.pm b/URI/http.pm
index cb69822..8631e40 100644
--- a/URI/http.pm
+++ b/URI/http.pm
@@ -13,7 +13,7 @@ sub canonical
     my $other = $self->SUPER::canonical;

     my $slash_path = defined($other->authority) &&
-        !length($other->path) && !defined($other->query);
+        !length($other->path) && defined($other->query);

     if ($slash_path) {
        $other = $other->clone if $other == $self;
diff --git a/t/http.t b/t/http.t
index 2b8c44b..65e32cd 100644
--- a/t/http.t
+++ b/t/http.t
@@ -1,6 +1,6 @@
 #!perl -w

-print "1..15\n";
+print "1..17\n";

 use URI;

@@ -61,3 +61,12 @@ $u = URI->new("http://%77%77%77%2e%70%65%72%6c%2e%63%6f%6d/%7
 print "not " unless $u->canonical eq "http://www.perl.com/pub/a/2001/08/27/bjor
 print "ok 15\n";

+# RFC 1738: "/" can't be ommited when <searchpart> is present.
+$u = URI->new("http://localhost?p=1");
+print "not " unless $u->canonical eq "http://localhost/?p=1";
+print "ok 16\n";
+
+$u = URI->new("http://localhost");
+print "not " unless $u->canonical eq "http://localhost";
+print "ok 17\n";
+

[gisle]

RFC 2396 does not agree on this AFAICT.

[hiratara]

RFC3986 says that RFC2396 was obsolete. So we should follow RFC3986(and RFC1738), shouldn't we ??

http://www.ietf.org/rfc/rfc3986.txt

[gisle]

RFC3986:

URI           = scheme ":" hier-part [ "?" query ] [ "#" fragment ]
hier-part     = "//" authority path-abempty
path-abempty  = *( "/" segment )

This does allow "http://localhost?p=1". Does compatiblity with other stuff break with this, or do you want to argue the aesthetics of of this as canonical form?

[gisle]

We still have this inconsistency:

$ perl -MURI -le 'print URI->new("http://foo#bar")->canonical'
http://foo/#bar
$ perl -MURI -le 'print URI->new("http://foo?bar")->canonical'
http://foo?bar

It's better to be consistent, so one of these ought to change.

[Copilot]

Pull Request Overview

This PR fixes the canonical URL formatting to comply with RFC 1738, which requires the "/" path separator to be retained when a query string is present. The change ensures that URLs like "http://localhost?p=1" are canonicalized to "http://localhost/?p=1" instead of omitting the slash.

• Modifies the canonical method logic to always add "/" when path is empty, regardless of query presence
• Adds a test case to verify the fix works correctly for URLs with query strings

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
URI/http.pm	Simplified slash_path condition to always add "/" when path is empty and authority is defined
t/http.t	Added test case for URL canonicalization with query string and updated test count

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

There's a typo in the comment: 'ommited' should be 'omitted'.

Suggested change

	# RFC 1738: "/" can't be ommited when <searchpart> is present.
	# RFC 1738: "/" can't be omitted when <searchpart> is present.