Use bindings to the C++ mbgl

[josxha]

Description

The package currently uses the MapLibre Native SDKs for Android and iOS and displays the map using Plattform Views. This solution is used by Google and Mapbox as well and is a relatively simple integration.
Flutter (Dart) -> MapLibre Native Android (Kotlin) -> MBGL (C++)

It is possible to integrate the MBGL renderer directly in Flutter using the Texture API. There is a Medium Article has gives some additional guidance.
Platform views are only available on iOS and Android. When directly using the C++ part, the package could be used on linux, macos and windows as well.

Similar Attempts

• There has been a previous attempt to integrate mbgl for flutter_map: Source Code, Slack Thread.
• A company wants to add Flutter bindings to galileo, a new mapping library written in rust: Slack Thread, GitHub Comment
• Another attempt to write a new renderer using flutter_gpu on impeller: Source Code, Slack Thread
• The oldest attempt is a renderer that is written entirely in Dart: Source Code

In my opinion all attempts are great to see. Having this thread for the mbgl renderer doesn't mean that the package couldn't use an other solution.

[gabbopalma]

I think having a direct implementation with the MapLibre C++ engine would be an amazing thing.
But looking at the differences, with the current solution via Platform Views for Android and iOS, would there be significant changes in performance?
Besides the possible implementation of Flutter MapLibre on MacOS, Windows and Linux, what would be the biggest benefits of this implementation?
Would it lead to possible more customized solutions if requested by potential Partners of the package?

[josxha]

would there be significant changes in performance?

There should be a performance improvement but it's hard to tell how significant it is. It might have similar performance to platform views that use the texture layer mode but kicks out an abstraction layer.

Besides the possible implementation of Flutter MapLibre on MacOS, Windows and Linux, what would be the biggest benefits of this implementation?

Ideally, rendering of the map would be in sync with the rendering of the Flutter app. While this is straight off the table when we use platform views there might be a chance here. But we don't have a POC, so it's currently just speculation.

Would it lead to possible more customized solutions if requested by potential Partners of the package?

It depends on the particular requirements but in general I'd say yes. Direct access to mbgl would mean that Cache, Offline and Location would all move to Dart and would allow better access.

[gabbopalma]

If there is a way to test the implementation of core features (drawing features on map and gestures) and do a stress test with these things, I would say it won't hurt to try.

It would be very interesting to be independent of the Android and iOS packages, making the Flutter package very competitive for future Partners as well.