Need help with appropriate usage of DataFilterExtension filter_categories

[ATL2001]

I'm not sure what I'm doing wrong, but I can't get the points on my map to filter using filter_categories.

this is the code I'm using to make my data and layer with the data filter extension add it to my Lonboard map

import geopandas as gpd
import ipywidgets 
from shapely.geometry import Point

import lonboard
from lonboard.basemap import CartoBasemap
from lonboard.layer_extension import DataFilterExtension

d = {
    'col_to_filter': [0.0, 1.0, 2.0, 3.0, 4.0, 5.0], 
    'geometry': [
        Point(0, 0), 
        Point(1, 1), 
        Point(2, 2), 
        Point(3, 3), 
        Point(4, 4), 
        Point(5, 5)
    ]
}
gdf = gpd.GeoDataFrame(d, crs="EPSG:4326")

dfe = DataFilterExtension(filter_size=1, category_size=1)
point_layer = lonboard.ScatterplotLayer.from_geopandas(
    gdf,
    get_fill_color=(0,255,0),
    radius_min_pixels=10,
    extensions=[dfe],
    get_filter_category=gdf["col_to_filter"]
)

m = lonboard.Map(layers=[point_layer], basemap_style=CartoBasemap.DarkMatter)
ipywidgets.VBox([m])

and then after I create the map, I was thinking that this should result in only the point at (2,2) being displayed, but when I run it, they're all still there :(

point_layer.filter_categories = [2]

can anyone help?

[kylebarron]

I think the underlying deck.gl filter categories is more meant for non-float data. It might be an issue with float comparisons here, especially if 2 is uploaded once to the GPU as a float32 and maybe another time as an integer, or something like that.

If you want to choose a specific category that's based on a float value, perhaps first consider a min range and max range that are within an epsilon of that value.

Maybe try filter_categories with a string value? If we get a working example we should add it to the docs, because I know it's a little finnicky

[ATL2001]

yeah, I'll certainly make up an example notebook if I get it working! I tried using strings after looking at the deck.gl docs but when I pass that column from my dataframe to the get_filter_category parameter I end up with a TraitError.

TraitError: The 'get_filter_category' trait of a ScatterplotLayer instance expected numeric dtype, not the ndarray array(['0', '1', '2', '3', '4', '5'], dtype=object)

I may tinker around this weekend with the validate funciton in traits.py to not check for numeric dtypes and see if I can get it to work with strings.

[kylebarron]

That sounds awesome, thanks 🙏