Add method to convert a URI string to a uri object?

[pdehaan]

I've always been a bit confused by this uri object thing:

var uri = {
  spec: "http://fakehost/test/page.html",
  host: "fakehost",
  prePath: "http://fakehost",
  scheme: "http",
  pathBase: "http://fakehost/test"
};
var result = new window.Readability(uri, window.document).parse();

It seems that all the fields can be extracted from the spec string. It'd be nice if there were just some helper function which converts a fully qualified URL into a host, prePath, scheme, and pathBase for you.

Or maybe modify the window.Readability() method to accept an object or string, and if it's a URI string, try and auto-convert it to an object.

I'm guessing this would be pretty trivial w/ Node.js' url.parse() function, but it may fit better in the Readability.js library as plain olde JavaScript.

var url = require('url');

function uriFromString(str) {
  var urlObj = url.parse(str, true);
  var protocol = urlObj.protocol.replace(':', '');
  var prePath = protocol + '://' + urlObj.hostname;
  var uri = {
    spec: urlObj.href,
    host: urlObj.hostname,
    pathBase: prePath + urlObj.pathname.replace(/[^\/]+$/, ''),
    prePath: prePath,
    scheme: protocol
  };
  return uri;
}

console.log(JSON.stringify(uriFromString(uri.spec), null, 2));

I think we've also used http://medialize.github.io/URI.js/ in other projects (via Bower).

[gijsk]

URI parsing is surprisingly treacherous ground (cf. https://bugzilla.mozilla.org/show_bug.cgi?id=483304 https://bugzilla.mozilla.org/show_bug.cgi?id=1093611 ). If the tests are the only place where we need to make one of these (because otherwise consumers are responsible for creating these), I'd be inclined to just leave it. OTOH, security/correctness concerns might be limited for test-only use as well?

Is there anything using this / some variant of this that isn't the tests?

(don't care about the real code reading the object - worried about how they're getting created though...)