Mojo JS bindings: change module loading solution.

mojo/public/js/new_bindings/base.js

Index: mojo/public/js/new_bindings/base.js
diff --git a/mojo/public/js/new_bindings/base.js b/mojo/public/js/new_bindings/base.js
new file mode 100644
index 0000000000000000000000000000000000000000..31173957a4790d0bf3e23e16740116a4d0900b10
--- /dev/null
+++ b/mojo/public/js/new_bindings/base.js
@@ -0,0 +1,31 @@
+// Copyright 2017 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+if (mojoBindings) {
+  throw new Error("mojoBindings has been initialized.");
+}
+
+var mojoBindings = {};

[Ken Rockot(use gerrit already), 2017/03/20 15:27:20]

I wonder about naming here, but it's kind of just a bikeshed. Should it be
MojoBindings, or do we reserve that naming style for IDL generated types? And I
suppose Mojo.Bindings is out of the question since that means monkeypatching a
generated namespace.

[yzshen1, 2017/03/20 17:45:02]

According to JS style guide, namespace should be namedLikeThis (with lower case
initial).
However, we have things like "Mojo.createMessagePipe" in the WebIDL-generated
API. I think we view "Mojo" as a type instead of a namespace in this case. A
supporting evidence is that we are not using Mojo.Handle, instead we use
MojoHandle.

Do you think we need to start a bikeshedding thread for naming of the
WebIDL-generated API? :) It is usually not used directly so it shouldn't be too
much work even if we would like to change the names later.

Another thing I would like to bring up is whether we want to do our best to
convert mojom names to match JS coding style?
At the moment, method names are converted to match JS style.

However we don't do it for most names (module name, field name, parameter name,
etc.) For example we map

module "some_namespace";

to a namespace "some_namespace" in JS, that is not the format recommended by JS
style guide ("someNamespace").

IMO, it would be nice to follow the style guide as close as possible. And it
probably makes sense to do it at the time while we are switching to a "new
mode", although the work is a little tedious for me. :)

WDYT?

[Ken Rockot(use gerrit already), 2017/03/21 16:42:43]

I agree with this rationale, so SGTM. In fact I was going to make the same
suggestion about module namespaces.

[alokp, 2017/03/27 01:56:25]

Bikeshed: I find mojoBindings ugly :)

I am open to installing the Mojo object into window.mojo and then we can extend
the mojo object with everything in the Bindings layer? Then consumers of the API
do not need to care about the distinction between core and bindings? This would
also be consistent with C++ API where everything is in the mojo namespace.

+mojoBindings.internal = {};
+mojoBindings.internal.global = this;
+
+(function() {
+  var internal = mojoBindings.internal;
+
+  function exposeNamespace(namespace) {
+    var current = internal.global;
+    var parts = namespace.split('.');
+
+    for (var part; parts.length && (part = parts.shift());) {
+      if (!current[part]) {
+        current[part] = {};
+      }
+      current = current[part];
+    }
+
+    return current;
+  }
+
+  internal.exposeNamespace = exposeNamespace;
+})();