Slash automatically added to URI with fragment and no path

[jamesamcl]

> URI('http://example.com#foo').toString()
'http://example.com/#foo'

As far as I know, http://example.com#foo is a valid URI. This is quite a big issue for me because if the URI changes after loading/serialization, it completely invalidates my RDF.

[jamesamcl]

Related to #221.

[jamesamcl]

Seems you're right as far as normalisation guidelines go for HTTP - RFC 3986 says:

In general, a URI that uses the generic syntax for authority with an
empty path should be normalized to a path of "/".  Likewise, an
explicit ":port", for which the port is empty or the default for the
scheme, is equivalent to one where the port and its ":" delimiter are
elided and thus should be removed by scheme-based normalization.  For
example, the second URI above is the normal form for the "http"
scheme.

However, this is HTTP specific and shouldn't happen for URIs with other protocols.

[rodneyrehm]

You are correct, http://example.com#foo is a valid URI, as path-abempty may indeed be the empty string (per RFC 3986 Collected ABNF).

toString() runs URI.build() behind the scene, which adds the path if required:

var u = URI('http://example.com#foo');
u.toString()
// "http://example.com/#foo"

u._parts.path = null;
u.toString()
// "http://example.com#foo"

The slash is added by URI.parseAuthority() in case there is no path. Replacing "/" with "" (empty string) would not help, as URI.build() would interpret the empty string to mean /. However, returning null instead of the empty string would solve your problem - and make 13 tests fail, thus has to be considered a breaking change.

However, this is HTTP specific and shouldn't happen for URIs with other protocols.

I agree, but your example showed an http url?

[jamesamcl]

Thanks,

I agree, but your example showed an http url?

Yes - I only found out about the difference with HTTP afterwards.

Maybe permitting empty paths could be an option to avoid breaking tests?

[rodneyrehm]

Maybe permitting empty paths could be an option to avoid breaking tests?

how do you suggest we do that?