In [1]:

# /// script
# requires-python = ">=3.12"
# dependencies = [
#     "geodatasets",
#     "geopandas",
#     "ipywidgets",
#     "lonboard",
#     "numpy",
#     "palettable",
#     "pandas",
#     "shapely",
# ]
# ///

# /// script # requires-python = ">=3.12" # dependencies = [ # "geodatasets", # "geopandas", # "ipywidgets", # "lonboard", # "numpy", # "palettable", # "pandas", # "shapely", # ] # ///

Categorical Filtering with the DataFilterExtension¶

The DataFilterExtension adds GPU-based data filtering functionalities to layers, allowing the layer to show/hide objects based on user-defined properties.

Dependencies¶

Install uv and then launch this notebook with:

uvx juv run examples/data-filter-extension-categorical.ipynb

(The uvx command is included when installing uv).

Categorical Filtering¶

In this example the DataFilterExtension will be used to filter the display the home sales dataset from geodatasets based on the number of bedrooms, bathrooms.

To demonstrate, we'll create:

1. A GeoPandas GeoDataFrame of home sales data.
2. A Lonboard ScatterPlotLayer from the GeoDataFrame that has a DataFilterExtension set up for categorical filtering.
3. Some IPyWidgets linked to the ScatterPlotLayer's filter_categories property to allow us to interactively filter the points on the map.

Imports¶

In [2]:

import geodatasets
import geopandas
import traitlets
from ipywidgets import Button, HBox, SelectMultiple, VBox
from palettable.colorbrewer.diverging import RdYlGn_11

import lonboard
from lonboard import Map, ScatterplotLayer
from lonboard.layer_extension import DataFilterExtension

import geodatasets import geopandas import traitlets from ipywidgets import Button, HBox, SelectMultiple, VBox from palettable.colorbrewer.diverging import RdYlGn_11 import lonboard from lonboard import Map, ScatterplotLayer from lonboard.layer_extension import DataFilterExtension

Reading the Home Sales Data¶

This example makes use the geodatasets python package to access some spatial data easily.

Calling geodatasets.get_path() will download data the specified data to the machine and return the path to the downloaded file. If the file has already been downloaded it will simply return the path to the file. See downloading and caching for further details.

In [3]:

home_sales_df = geopandas.read_file(geodatasets.get_path("geoda.home_sales"))[
    ["price", "bedrooms", "bathrooms", "geometry"]
]
home_sales_df

home_sales_df = geopandas.read_file(geodatasets.get_path("geoda.home_sales"))[ ["price", "bedrooms", "bathrooms", "geometry"] ] home_sales_df

Out[3]:

	price	bedrooms	bathrooms	geometry
0	221900.0	3.0	1.0	POINT (-122.257 47.5112)
1	538000.0	3.0	2.0	POINT (-122.319 47.721)
2	180000.0	2.0	1.0	POINT (-122.233 47.7379)
3	604000.0	4.0	3.0	POINT (-122.393 47.5208)
4	510000.0	3.0	2.0	POINT (-122.045 47.6168)
...	...	...	...	...
21608	360000.0	3.0	2.0	POINT (-122.346 47.6993)
21609	400000.0	4.0	2.0	POINT (-122.362 47.5107)
21610	402101.0	2.0	0.0	POINT (-122.299 47.5944)
21611	400000.0	3.0	2.0	POINT (-122.069 47.5345)
21612	325000.0	2.0	0.0	POINT (-122.299 47.5941)

21613 rows × 4 columns

Create the ScatterplotLayer with a DataFilterExtension extension for categorical filtering¶

The DataFilterExtension will be created with filter_size=None to indicate we do not want to use a range filter, and category_size=2 to indicate we want to use two different categories from the data to filter the data with explicit values.

The points in the layer will be symbolized based on a continuous colormap of the price. Lower priced homes will be red, and higher priced homes will be green. We'll throw out the upper and lower 5% of value from the color map so the upper and lower outliers do not influence the colormap.

In [4]:

min_bound = home_sales_df["price"].quantile(0.05)
max_bound = home_sales_df["price"].quantile(0.95)
price = home_sales_df["price"]
normalized_price = (price - min_bound) / (max_bound - min_bound)

home_sale_layer = ScatterplotLayer.from_geopandas(
    home_sales_df,
    get_fill_color=lonboard.colormap.apply_continuous_cmap(normalized_price, RdYlGn_11),
    radius_min_pixels=5,
    extensions=[
        DataFilterExtension(filter_size=None, category_size=2),
    ],
    get_filter_category=home_sales_df[["bedrooms", "bathrooms"]].values,
    filter_categories=[[], []],
)

min_bound = home_sales_df["price"].quantile(0.05) max_bound = home_sales_df["price"].quantile(0.95) price = home_sales_df["price"] normalized_price = (price - min_bound) / (max_bound - min_bound) home_sale_layer = ScatterplotLayer.from_geopandas( home_sales_df, get_fill_color=lonboard.colormap.apply_continuous_cmap(normalized_price, RdYlGn_11), radius_min_pixels=5, extensions=[ DataFilterExtension(filter_size=None, category_size=2), ], get_filter_category=home_sales_df[["bedrooms", "bathrooms"]].values, filter_categories=[[], []], )

Create the iPyWidgets to interact with the DataFilterExtension¶

Since we want to only display the points which are tied to specific number of bedrooms and bathrooms we'll:

1. Create two ipywidgets SelectMultiple widgets which will hold the different numbers of bedrooms and bathrooms in the home sales data.
2. Create two ipywidgets Button widgets which will clear the selected values for the bedrooms and bathrooms.
3. Observe the changes made to the SelectMultiple widgets to update the layer's filter_categories property.

This will enable us to select one or more number of bedrooms or bathrooms and have the map instantly react to display only the data that matches the selections. If a select widget does not have a selection, all the values from that selector will be used.

In [5]:

unique_bedrooms_values = list(home_sales_df["bedrooms"].sort_values().unique())
unique_bathrooms_values = list(home_sales_df["bathrooms"].sort_values().unique())

bedrooms_select = SelectMultiple(description="Bedrooms", options=unique_bedrooms_values)
bathrooms_select = SelectMultiple(
    description="Bathrooms",
    options=unique_bathrooms_values,
)


def on_select_change(_: traitlets.utils.bunch.Bunch = None) -> None:
    """Set the layer's filter_categories property based on widget selections."""
    bedrooms = bedrooms_select.value
    bathrooms = bathrooms_select.value
    if len(bedrooms) == 0:
        bedrooms = unique_bedrooms_values
    if len(bathrooms) == 0:
        bathrooms = unique_bathrooms_values
    home_sale_layer.filter_categories = [bedrooms, bathrooms]


bedrooms_select.observe(on_select_change, "value")
bathrooms_select.observe(on_select_change, "value")

clear_bedrooms_button = Button(description="Clear Bedrooms")


def clear_bedrooms(_: Button) -> None:
    bedrooms_select.value = []


clear_bedrooms_button.on_click(clear_bedrooms)

clear_bathrooms_button = Button(description="Clear Bathrooms")


def clear_bathrooms(_: Button) -> None:
    bathrooms_select.value = []


clear_bathrooms_button.on_click(clear_bathrooms)

home_sale_map = Map(
    layers=[home_sale_layer],
    basemap=lonboard.basemap.MaplibreBasemap(),
)
on_select_change()  # fire the function once to initially set the layer's filter_categories, and display all points

display(home_sale_map)
display(
    HBox(
        [
            VBox([bedrooms_select, clear_bedrooms_button]),
            VBox([bathrooms_select, clear_bathrooms_button]),
        ],
    ),
)

unique_bedrooms_values = list(home_sales_df["bedrooms"].sort_values().unique()) unique_bathrooms_values = list(home_sales_df["bathrooms"].sort_values().unique()) bedrooms_select = SelectMultiple(description="Bedrooms", options=unique_bedrooms_values) bathrooms_select = SelectMultiple( description="Bathrooms", options=unique_bathrooms_values, ) def on_select_change(_: traitlets.utils.bunch.Bunch = None) -> None: """Set the layer's filter_categories property based on widget selections.""" bedrooms = bedrooms_select.value bathrooms = bathrooms_select.value if len(bedrooms) == 0: bedrooms = unique_bedrooms_values if len(bathrooms) == 0: bathrooms = unique_bathrooms_values home_sale_layer.filter_categories = [bedrooms, bathrooms] bedrooms_select.observe(on_select_change, "value") bathrooms_select.observe(on_select_change, "value") clear_bedrooms_button = Button(description="Clear Bedrooms") def clear_bedrooms(_: Button) -> None: bedrooms_select.value = [] clear_bedrooms_button.on_click(clear_bedrooms) clear_bathrooms_button = Button(description="Clear Bathrooms") def clear_bathrooms(_: Button) -> None: bathrooms_select.value = [] clear_bathrooms_button.on_click(clear_bathrooms) home_sale_map = Map( layers=[home_sale_layer], basemap=lonboard.basemap.MaplibreBasemap(), ) on_select_change() # fire the function once to initially set the layer's filter_categories, and display all points display(home_sale_map) display( HBox( [ VBox([bedrooms_select, clear_bedrooms_button]), VBox([bathrooms_select, clear_bathrooms_button]), ], ), )

VBox(children=(<lonboard._map.Map object at 0x1127263f0>, VBox(children=(ErrorOutput(), ErrorOutput()), layout…

HBox(children=(VBox(children=(SelectMultiple(description='Bedrooms', options=(np.float64(0.0), np.float64(1.0)…

In [ ]: