Sun calculator

[Gustl22]

Sun calculator with accuracy of ~1 minute.
Can test actual sun position with: BuildingLayer.getExtrusionRenderer().enableCurrentSunPos(true);

Or to see the whole day cycle use this in MapsforgeTest.createLayers():

BuildingLayer extrusionLayer;
if (s3db) {
    extrusionLayer = new S3DBLayer(mMap, l);
} else
    extrusionLayer = new BuildingLayer(mMap, l);

final ExtrusionRenderer extrusionRenderer = extrusionLayer.getExtrusionRenderer();

mMap.layers().add(extrusionLayer);

extrusionRenderer.enableCurrentSunPos(false);
mMap.events.bind(new Map.UpdateListener() {
    @Override
    public void onMapEvent(Event e, MapPosition mapPosition) {
        long t = System.currentTimeMillis();
        float progress = (((t % 2000) / 1000f));

        extrusionRenderer.getSun().setProgress(progress);
        extrusionRenderer.getSun().updatePosition();
        extrusionRenderer.getSun().updateColor(); // only relevant for shadow implementation

        mMap.updateMap(true);
    }
});

See #575.

[devemux86]

Quick comment: since the feature is optional, better have a lazy initialization of any feature objects, specially if they carry collections.

[Gustl22]

You mean lazy init the sun, or the collection?
Probably can achieve lazy init collection when calling updateColor() the first time.
Yes it's a feature, but fully customizable. Storing in another attribute next to the sun would be confusing.

[devemux86]

The Sun class.
Can use getSun everywhere and if null at first call, just create it there.

[Gustl22]

But we have to specify light position in any case?

[devemux86]

Aren't shadows optional?

[Gustl22]

All parameters in sun are optional. So why make an option of an option, which in fact gives the same results and only makes it more complicated.

Yes this whole construct is needed for shadows, too. But also gives new options for extrusion renderer only.

[Gustl22]

I don't think storing 5 values more has impact on performance or memory. It's a more physical based way to specify lights position.

[devemux86]

Let's wait to be reviewed then. Most concern is with users who simply don't care for light or shadows and want to continue with same fast rendering without new possible GL issues.

[Gustl22]

I don't see any regression here. If somebody don't care for new features, can stay on tested versions.

[devemux86]

I don't see any regression here.

Regressions can be seen with extensive testing. 🙂

If somebody don't care for new features, can stay on tested versions.

That cannot be an answer to users: if don't need a feature, use previous lib version.
Proper answer is: can enable the new optional feature, if need via the XYZ method.

[Gustl22]

Regressions can be seen with extensive testing.

Of course, there always can be an issue. But in this case the position parameter only moved to Sun and will not change after initialization by default.

Proper answer is: can enable the new optional feature, if need via the XYZ method.

You're right, but not all implementations can be optional, only if they offer a different and reasonable result. And I added a bunch of methods to make it all customizable. So e.g. can set the position simply via setPosition.

BTW: I removed one TODO...already working.

[devemux86]

Thanks, merged via eb3efa7.

[devemux86]

I also added the above test in MapsforgeTest.

[Gustl22]

Nice thanks!