[android] Refactor annotations API

[ljbade]

Remove unimplemented properties.
Make annotations immutable.
Correct defintions of equals() and hasCode().
Change anchor U/V to int.
Move .alpha to MultiPoint and anchor() to PolylineOptions and PolygonOptions.
Make InfoWindow classes package private.
Add setOnInfoWindowClickListener and remove old method from Marker.
JavaDoc internal methods with "Do not use."
Refactor showInfoWindow() to remove need for exposing internal method.
Make select/deselectMarker public. Add getSelectedMarker.
Fix bug where you couldn't reselect a closed info window.
Fixes #2546
Fixes #2448

[ljbade]

There is a lot going on in this commit, but the result is an annotations API that only exposes what we have implemented. It also improves parts of the code, and fixes a bug that had not been ticketed before.

I will aim to get this merged tonight, ready for Sirius build tomorrow.

I have carefully tested everything with the TestApp and there should be no new bugs introduced.

/cc @bleege @tobrun

[ljbade]

The only Travis failure was the known hang in the iOS tests. Restarting that job, if it still fails we can merge as this is an Android only change.

[tobrun]

@ljbade
When doing a couple of test runs I was able to produce a crash.
I'm not able to reproduce it anymore, I believe I was too fast to press a marker before the infowindow was ready to handle the click event. This could well be a crash related to my code an not this PR. If you haven't picked this up tonight I will later this day.

java.lang.NullPointerException: Attempt to invoke virtual method 'void com.mapbox.mapboxsdk.annotations.Marker.showInfoWindow()' on a null object reference
at com.mapbox.mapboxsdk.views.MapView.adjustTopOffsetPixels(MapView.java:1978)
at com.mapbox.mapboxsdk.views.MapView.access$500(MapView.java:98)
at com.mapbox.mapboxsdk.views.MapView$1.onMapChanged(MapView.java:828)
at com.mapbox.mapboxsdk.views.MapView$3.run(MapView.java:2823)
at android.os.Handler.handleCallback(Handler.java:739)
at android.os.Handler.dispatchMessage(Handler.java:95)
at android.os.Looper.loop(Looper.java:148)
at android.app.ActivityThread.main(ActivityThread.java:5417)
at java.lang.reflect.Method.invoke(Native Method)
at com.android.internal.os.ZygoteInit$MethodAndArgsCaller.run(ZygoteInit.java:726)
at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:616)

[ljbade]

@tobrun That is a strange crash, I can't see how it gets a null since it checks for that specifically:

if (mSelectedMarker != null) {
            if (mSelectedMarker.isInfoWindowShown()) {
                mSelectedMarker.hideInfoWindow();
                mSelectedMarker.showInfoWindow();
            }
        }

[ljbade]

Ah might be race condition between the UI thread and the main thread. Map change event comes from main thread, other changes such as touch input come from UI thread.

[ljbade]

I think this crash is from changes in #2627

[ljbade]

No doesn't look like map change problem since we already forward them to the UI thread.

Something must be touching MapView from a non-UI thread when it shouldn't.

@tobrun If you can't reproduce again, I think we can chalk it up to your device playing weird. And merge.

[tobrun]

@tobrun If you can't reproduce again,
I think we can chalk it up to your device playing weird. And merge.

@ljbade 👍

[bleege]

@ljbade This is a LARGE change so close to launch. I'm not comfortable merging this into master right now. Let's bump this to android-v2.2.0 to give us more time.

[ljbade]

Sure, should we quickly add some do not use or not implemented javadoc to the annotation functions that don't do anything for 2.1.0 pending there removal?

[bleege]

should we quickly add some do not use or not implemented javadoc to the annotation functions that don't do anything for 2.1.0 pending there removal?

I was under the impression that these were going to be made private access? If they haven't already then let's do that.

[ljbade]

@bleege OK I'll make a version of this PR that includes just those changes (make non-functional methods private). I can't make all of the internal only methods private though (that was a big part of this refactor - to make it possible to make some of them private).

I'll include the .alpha fix too.

[ljbade]

@bleege The other thing I've been puzzling about is how Google keep the annotation classes in the map.model package which is separate from the GoogleMap object in map. They manage to do this without exposing any internal APIs.

Do they perhaps strip internal functions from the SDK before the publish? This is how they do it in the Android framework.

[ljbade]

Finally got this branch up to date with v2.1.0 merge.

Waiting for Travis.

[ljbade]

Only failing on iOS due to hang in location tests. Merging.