bluetooth: Use BluetoothUUID instead of string when sending uuids

device/bluetooth/mojo/bluetooth_uuid.typemap

Index: device/bluetooth/mojo/bluetooth_uuid.typemap
diff --git a/device/bluetooth/mojo/bluetooth_uuid.typemap b/device/bluetooth/mojo/bluetooth_uuid.typemap
new file mode 100644
index 0000000000000000000000000000000000000000..efdc0a6df71a6bb5103d7cada3ff7227c72eb6e0
--- /dev/null
+++ b/device/bluetooth/mojo/bluetooth_uuid.typemap
@@ -0,0 +1,11 @@
+# Copyright 2016 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.
+
+mojom = "//device/bluetooth/mojo/bluetooth_uuid.mojom"
+public_headers = [ "//device/bluetooth/bluetooth_uuid.h" ]
+traits_headers = [ "//device/bluetooth/mojo/bluetooth_uuid_struct_traits.h" ]
+deps = [
+  "//device/bluetooth",
+]
+type_mappings = [ "device.mojom.BluetoothUUID=std::unique_ptr<device::BluetoothUUID>" ]

[ortuno, 2016/05/26 20:46:13]

It would be nice if I could define a different type that is used for nullable
fields e.g.:

type_mappings = [ "device.mojom.BluetoothUUID=device::BluetoothUUID" ]
type_mappings = [
"device.mojom.BluetoothUUID?=std::unique_ptr<device::BluetoothUUID>" ]

[Ken Rockot(use gerrit already), 2016/05/26 21:05:07]

I don't think we want to force people to do that. I think we should try to
establish a general purpose solution for nullable types that we can apply
regardless of typemapping. This is an ongoing discussion that we can hopefully
resolve soon.

As an alternative interim solution for you, would it maybe be cleaner to support
an "invalid" state for your device::BluetoothUUID type rather than map to a
unique_ptr?

[ortuno, 2016/05/27 14:38:35]

Yes. That would be even nicer.
BluetoothUUID already has an "invalid" state but we use that for validation. See
bluetooth_uuid_struct_traits.h

[Ken Rockot(use gerrit already), 2016/05/27 14:57:42]

Doh, right.