switch mediaGalleries to .idl api definition

chrome/common/extensions/api/experimental_media_galleries.idl

Index: chrome/common/extensions/api/experimental_media_galleries.idl
diff --git a/chrome/common/extensions/api/experimental_media_galleries.idl b/chrome/common/extensions/api/experimental_media_galleries.idl
new file mode 100644
index 0000000000000000000000000000000000000000..ae42828bc94c6b08b75b65c5cd872089906e73c4
--- /dev/null
+++ b/chrome/common/extensions/api/experimental_media_galleries.idl
@@ -0,0 +1,53 @@
+// Copyright (c) 2012 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.
+
+// File-level comment to appease parser. Eventually this will not be necessary.
+
+namespace experimental.mediaGalleries {
+
+  enum GetMediaFileSystemsInteractivity {
+    silent, prompt, prompt_if_needed
+  };
+
+  [inline_doc] dictionary MediaFileSystemsDetails {
+    // Whether to prompt the user for additional media galleries before
+    // returning the permitted set. Default is silent.
+    GetMediaFileSystemsInteractivity? interactivity;
+  };
+
+  dictionary AssembleMediaFileDetails {
+    // TODO(estade): this should be [instanceOf=Blob].
+    object mediaFileContents;
+
+    // TODO(estade): this should be [instanceOf=Metafile].
+    object metadata;
+  };
+
+  // TODO(estade): the type of the argument should be LocalFileSystem[].

[vandebo (ex-Chrome), 2012/07/19 19:58:31]

Do you want to include bug numbers here?

[Evan Stade, 2012/07/19 20:55:28]

are there bug numbers for this? I'm not even sure there's any bug preventing it.
[instanceOf] works, I just don't know if it supports arrays, I figured I'd look
into that later.

[vandebo (ex-Chrome), 2012/07/19 22:45:22]

Sorry, not sure.  I initially assumed it was the same issue as the bug comments
in the .json.

[Evan Stade, 2012/07/19 23:06:33]

no, i believe that bug's resolved because instanceOf does work.

+  callback MediaFileSystemsCallback = void (object mediaFileSystems);

[vandebo (ex-Chrome), 2012/07/19 19:58:31]

mediaFileSystems => arrayOfMediaFileSystems ?

[Evan Stade, 2012/07/19 20:55:28]

the type tells you "arrayOf", and the trailing "s" also implies it.

[vandebo (ex-Chrome), 2012/07/19 22:45:22]

What do you mean "the type" tells you...?  Not sure why you want to fight about
a clarifying name change.

[Evan Stade, 2012/07/19 23:06:33]

By "the type" I mean once the TODO is fixed, there will be a type which tells
you the parameter is an array. I believe variable names should make sense, but
should not include types. Hungarian notation is explicitly disallowed by the C++
style guide.

[vandebo (ex-Chrome), 2012/07/20 18:30:26]

I see... Is there something hard about fixing this todo (something more than
changing the text here?).  If so, can you do array of object in the meantime?

[Evan Stade, 2012/07/20 19:48:12]

yes, the hard thing is figuring out the syntax. After reading through the idl
parser I think I got the correct thing now though.

+  callback AssembleMediaFileCallback =
+      void ([instanceOf=Blob] optional object mediaFile);
+
+  interface Functions {
+    // Get the media galleries configured in this user agent. If none are
+    // configured or available, the callback will receive an empty array.
+    static void getMediaFileSystems(optional MediaFileSystemsDetails details,

[vandebo (ex-Chrome), 2012/07/19 19:58:31]

Does this mean that the interactivity argument is optional? If so, optional
arguments should only go at the end... I don't think it needs to be optional.

[Evan Stade, 2012/07/19 20:55:28]

yes, it is optional, as I think it should be. Optional arguments need not go at
the end.

[vandebo (ex-Chrome), 2012/07/19 22:45:22]

Optional args not at the end of the arg list weirds me out, but if that's common
and acceptable in js, then so be it.

+                                    MediaFileSystemsCallback callback);
+
+    // Create a new MediaFile setting the metadata in the Blob to the supplied
+    // values, overriding any existing metadata in the media file. If user agent
+    // does not recognize the Blob as a supported file format, it will fail.
+    static void assembleMediaFile(AssembleMediaFileDetails details,
+                                  AssembleMediaFileCallback callback);
+
+    // Get any thumbnails contained in the passed media file. The resulting
+    // directory reader refers to a virtual directory that can not be navigated
+    // to. If there are no thumbnails in the passed file entry, the virtual
+    // directory will have no entries.
+    // TODO(estade): The return type should be Directory. The argument type
+    // should be [instanceOf=FileEntry].

[Evan Stade, 2012/07/19 20:55:28]

the reason I didn't include this instanceOf is not that instanceOf doesn't work,
but because it does work, and it beraks the tests (they just use dummy objects).

[vandebo (ex-Chrome), 2012/07/19 22:45:22]

Does it work properly?  I had locally set up the tests to use the right object
types but there were bugs that made that not work, so I pulled to get back to
simpler test code.

[Evan Stade, 2012/07/19 23:06:33]

yes, I had this and other types specified, but then the tests failed because
they were using {}.
I had hoped we'd be able to fix the tests and then define the types.

[vandebo (ex-Chrome), 2012/07/20 18:30:26]

You can use WebKitBlobBuilder to create a Blob test object.

[Evan Stade, 2012/07/20 19:48:12]

Done.

+    [nocompile] static object extractEmbeddedThumbnails(object mediaFile);

[vandebo (ex-Chrome), 2012/07/19 19:58:31]

Does nocompile mean that it's currently not implemented?  If so, we can do that
for assembleMediaFile too.

[asargent_no_longer_on_chrome, 2012/07/19 20:40:12]

nocompile means we don't autogenerate C++ code for method dispatch between the
renderer and browser processes, but the method signature is still exposed to JS.
Typically this is used for methods that have some or all of their implementation
in JS (typically in a file in chrome/renderer/resources/extensions/..).
 

[Evan Stade, 2012/07/19 20:55:28]

what Antony said, but as you know it's not implemented in js, it's implemented
in c++ in the renderer.

+  };
+
+};