Modifying URI should update all _parts? about uri.js

[rodneyrehm]

What you want to achieve is converting something that is parsed as a relative URI (because protocol is missing) to an absolute URI. The way you described is not the way to go about this, as modifying the "atomic component" protocol would also alter other components (hostname, port, path, …).

I would prefer something along the lines of

URI.absoluteUri = function(uri) {
  if (!uri.match(/^[a-z][a-z0-9-+-]*:\/\//i)) {
    uri = "http://" + uri;
  }

  return URI(uri);
};

URI.absoluteUri('ExAmPle.Org:80').toString();

I'm not sure if this needs to be a core function, though. Are you performing any other operations to sanitize user input?

[Jellyfrog]

Maybe a rebuild() function, but at the same time this feels a bit weird too.
Users might get confused since _string looks OK (after protocol('http')) but _parts still look wrong,
This maybe should be stated in the docs?

I was thinking that I could use URI.js to sanitize/validate
What I want do is something like the following:

var uri = new URI('ExAmPle.Org:80/a/b/c.htm?a=b#foo');
if(uri.protocol() != "http" && uri.protocol() != "https") {
  uri.protocol('http');
}

uri.fragment('');
uri.normalize();

if(!uri.protocol() || !uri.hostname() || !uri.path()) {
  //Url could not be fixed / validated
  return false;
}

[rodneyrehm]

While http:///some-path doesn't make sense, file:///some-path does. "foo.com" may be a directory, not necessarily a domain. So there can't be any protocol() magic. That means we can't plug this into any of the existing functions.

You can do a manual re-parse any time though:

var uri = new URI('ExAmPle.Org:80/a/b/c.htm?a=b#foo');
if (!uri.protocol()) {
  uri.protocol('http');
}

if (!uri.protocol().match(/^https?$/i)) {
  return false;
}

uri.fragment('');
// force a re-parse
uri.href(uri.href());
uri.normalize();

if(!uri.protocol() || !uri.hostname() || !uri.path()) {
  //Url could not be fixed / validated
  return false;
}

[Jellyfrog]

👍
I went with a implementation like you suggested, must have missed href() yesterday when I looked in the docs.
Maybe there should be a part in the docs about "invalid URIs"?
Since (like I said in the first post) its a bit confusing that the _string is OK but _parts is wrong