wip: Geocoder control
This is complex enough that it's not going to make it into this release
- The Maplibre geocoder implementation doesn't provide any out of the box support for any provider
- It's hard to match up the deck.gl widget backends to the maplibre geocoder API
Closes #681
Closest to out of the box would be their example which uses Nominatim: https://maplibre.org/maplibre-gl-js/docs/examples/geocode-with-nominatim/
I would've ideally liked to provide a selection of a few geocoder providers that users could choose from, because the actual geocoding API calls should be able to integrate into each library's solution. I don't really want to add maplibre-only or deck-only support, because library that's rendering the map view should be just an implementation detail.
I would've ideally liked to provide a selection of a few geocoder providers that users could choose from, because the actual geocoding API calls should be able to integrate into each library's solution. I don't really want to add maplibre-only or deck-only support, because library that's rendering the map view should be just an implementation detail.
If you have an idea of which providers you want to support (for a start), we can explore possible options. It is quite hard to support multiple geocoding providers in the long term, when most of them are commercial, while it is not the main priority of library authors either. I don't normally use Nominatim as I would need to self-host to use it for autocomplete (due to their usage limit).
Maplibre geocoder design is still useful as a general framework for mapping disparate geocoding inputs to outputs. Do you think it is possible to support one or two providers, and allow users to define their own? Through JS code etc
Ideally I would choose the top N geocoding providers by usage (not sure what those are), including at least one free one, if possible.
I don't currently have a way in Lonboard for users to provide their own JS code. Going through Python for geocoding requests might be too slow