feature: add support to zoom into a specific layer

[ncclementi]

When plotting a Map with multiple layers, I'd like to have an option to zoom in into a specific layer. For example, in this blog https://ibis-project.org/posts/ibis-duckdb-geospatial/ I have some code that plots a point, two lines (over some roads), and the NYC streets, that looks like this

broad_station_layer = ScatterplotLayer.from_geopandas(
    broad_station_gdf, get_fill_color="blue", get_radius=5
)
sts_near_broad_layer = PathLayer.from_geopandas(
    sts_near_broad_gdf, get_color="red", opacity=0.4, get_width=2
)
streets_layer = PathLayer.from_geopandas(streets_gdf, get_color="grey", opacity=0.3)
m = Map(
    [
        broad_station_layer,
        sts_near_broad_layer,
        streets_layer,
    ],
    view_state={"longitude": -74.01066, "latitude": 40.7069, "zoom": 16} # need this to zoom into desired outcome
)
m

However, without specifying the view_state the map displays like this, and the first two layers get completely lost unless I manually zoom in. It would be great if I could zoom in a box around let's say the first two layers or one of them, without having to find what is the lat, lon and specific zoom value.

[kylebarron]

All layers already have bounding box and weighted centroid attributes that are inferred from the input data.

lonboard/lonboard/_layer.py

Lines 75 to 76 in 5560ead

	_bbox = Bbox()
	_weighted_centroid = WeightedCentroid()

Currently it's on the Map class to compute a default view state from the union of all input layers:

lonboard/lonboard/_map.py

Lines 445 to 447 in 5560ead

	@traitlets.default("view_state")
	def _default_initial_view_state(self):
	return compute_view(self.layers)

compute_view converts that bbox and weighted centroid to a view state.

lonboard/lonboard/_viewport.py

Lines 58 to 79 in 5560ead

	def compute_view(layers: List[BaseLayer]):
	"""Automatically computes a view state for the data passed in."""
	bbox, center = get_bbox_center(layers)
	# When no geo column is found, bbox will have inf values
	try:
	zoom = bbox_to_zoom_level(bbox)
	return {
	"longitude": center.x,
	"latitude": center.y,
	"zoom": zoom,
	"pitch": 0,
	"bearing": 0,
	}
	except OverflowError:
	return {
	"longitude": center.x or 0,
	"latitude": center.y or 0,
	"zoom": 0,
	"pitch": 0,
	"bearing": 0,
	}

I think possibly the cleanest API here is to add a default_view_state attribute onto each layer, which passes the BaseLayer._bbox and BaseLayer._weighted_centroid into compute_view.

Then you'd be able to do

layer = ScatterplotLayer(...)
layer2 = PolygonLayer(...)
map = Map([layer, layer2], view_state=layer.view_state)

That would also let you e.g. override only the zoom, by setting

new_view_state = layer.view_state
new_view_state["zoom"] = 4
map.view_state = new_view_state

I don't have a lot of spare time right now; would you be interested in making a PR?

[ncclementi]

I'll take a look, but at the moment I'm a bit swamped, but I saw you'll be at Scipy next month, it might be a good place to pair on it and chat also about Ibis + duckdb + lonboard.