feat(android): parity for keyboardframechanged and keyboardVisible

[m1ga]

Add Android parity for keyboardframechanged event and Ti.App.keyboardVisible to Ti.App:

var win = Ti.UI.createWindow();
var tf = Ti.UI.createTextField({width: 100});
var tf2 = Ti.UI.createTextField({width: 100, top: 0});
win.add([tf,tf2]);

Ti.App.addEventListener("keyboardframechanged", function(e){
	alert("keyboard visible: " + Ti.App.keyboardVisible);
	if (Ti.App.keyboardVisible) {
		alert("Size: " + e.keyboardFrame.width + " x " + e.keyboardFrame.height);
	}
});
win.addEventListener("open", function(e){
	alert("keyboard visible: " + Ti.App.keyboardVisible);
})
win.open();

[m1ga]

simplescreenrecorder-2023-01-21_13.43.59.mp4

[hansemannn]

Could this be changed to finally bring parity for https://titaniumsdk.com/api/titanium/app.html#properties_keyboardvisible and https://titaniumsdk.com/api/titanium/app.html#events_keyboardframechanged ?

[m1ga]

I can trigger that event and use the keyboardVisible property 👍
Not all fields will be filled in the event but at least it will be there for Android and you can check the keyboard status!

[hansemannn]

Yep, great parity already!

[m1ga]

It's not in the Ti.App. namespace

• Ti.App.keyboardVisible
• and Ti.App.addEventListener("keyboardframechanged")

[m1ga]

Not sure about some of the errors in the test suite. Don't think they are related to this PR.
e.g. I see

Expected to throw exception\\' undefined undefined' to be 'Unable to convert null'"

in error.test.js Native exception surfaced. It's using Ti.Geolocation.accuracy = null; to check for 'Unable to convert null'.
It says its returning undefined but when I just run that line in a test app I see

so the error is Unable to convert null 🤷

I've tested my app, your app, kitchensink and hyperloop examples and they look fine 👍

[hansemannn]

Still, the onKeyboardChanged callback seems to be always called now, causing more work on the main thread, even when not using the feature. This should be prevented prior merging this PR

[m1ga]

It shouldn't fire the event because it will go this route:

it's not calling fireEvent in onKeyboardChanged, it's calling a method in TiApplication, from there it will call fireEvent and check hasListener

Where do you see the main thread activity? I could add the event back to the window so the check is in the same file but then it wouldn't be parity

[hansemannn]

I meant the Java callback invocation and construct of the kroll dict, sorry for the confusion

[m1ga]

exposed the hasListener and added it to the if checks:

fec7cc7#diff-6f50b13a3f7b68de5485b38155383cdbd1ec15887ad7fd84573b826d79154790R772

[hansemannn]

Thanks for all the adjustments! Sorry for being so picky..but I think it's worth to double-check the calls before it's getting an issue in production. We should be ready to go here..if someone else can verify the fix, we can merge.

[hansemannn]

Code-review approved, waiting for functional testing.

[hansemannn]

@m1ga Any external testing done? Maybe the users that requested this can help. This can always be the way to fast-lane changes.

[m1ga]

Since this is a Android 11+ fix it wasn't useful for the user (Android 8.1 device) so I don't know the current status. I don't have any use for it so I didn't implement it in a client app at the moment

[m1ga]

Tested on Android 14 (Pixel 9) and Android 15 (Emulator) again, working fine. Event is fired and keyboardVisible value is correct