Sidecar¶
The goal of this notebook is to show how using Jupyter Sidecar can help to interactively explore and understand your data.
Note that sidecar only works in JupyterLab and will not work in other notebook frontends like Colab.
Imports¶
import geopandas as gpd
import ipywidgets
import matplotlib as mpl
import pandas as pd
from palettable.colorbrewer.diverging import BrBG_10
from sidecar import Sidecar
from lonboard import Map, ScatterplotLayer
from lonboard.colormap import apply_continuous_cmap
For this example, we'll use a dataset of abandoned cars in Chicago. Since this is a CSV file, we need to pass in the names of the X and Y columns into the "open options" to pass to OGR.
url = "https://geodacenter.github.io/data-and-lab/data/Abandoned_Vehicles_Map.csv"
gdf = gpd.read_file(
url,
engine="pyogrio",
use_arrow=True,
X_POSSIBLE_NAMES="Longitude",
Y_POSSIBLE_NAMES="Latitude",
KEEP_GEOM_COLUMNS="NO",
)
gdf = gdf.set_crs("EPSG:4326")
This dataset has some null values for geometries. We'll remove these first:
gdf = gdf[gdf.geometry != None]
Let's also remove empty values from another column to make later steps easier (when modifying a single map object with lonboard, you can't change the number of rows).
gdf = gdf[gdf["How Many Days Has the Vehicle Been Reported as Parked?"] != ""]
We'll now initialize the Sidecar object, which splits the JupyterLab screen in half, creating the output area for our widget.
sc = Sidecar(title='My Map')
Now let's create a map layer from our GeoDataFrame.
The default map height is 500 pixels, but we'd like to have a taller map because it's vertical. This API may change in the future.
layer = ScatterplotLayer.from_geopandas(gdf)
map_ = Map(layers=[layer])
Now we can render this inside the sidecar. You should see the map appear on the right side of your screen.
with sc:
display(map_)
It's great that we can see the map, but let's customize the rendering a bit. There's a column in the data titled
How Many Days Has the Vehicle Been Reported as Parked?
Let's use this column to visualize how long it takes to remove a car in different areas of the city.
This is a string column, so first let's convert it to a numeric value.
colname = "How Many Days Has the Vehicle Been Reported as Parked?"
gdf[colname] = pd.to_numeric(gdf[colname])
In order to apply a colormap onto our data, we need to normalize our values to the range of 0-1. Matplotlib offers several normalization options. See the tutorial and the matplotlib.colors module.
To start, we'll do a simple linear normalization, converting from 0-90 to 0-1. Any values below 0 will be clamped to 0, and any values above 90 will be clamped to 1.
normalizer = mpl.colors.Normalize(0, 90)
normalized_values = normalizer(gdf[colname])
These values now range between 0 and 1:
normalized_values
BrBG_10.mpl_colormap
layer.radius_min_pixels = 0.5
layer.get_fill_color = apply_continuous_cmap(normalized_values, BrBG_10)
Now the circles on the map have colors based on the time they've been parked.
But this dataset is skewed with a long tail. Most cars have been sitting way less than 90 days, but a few have been sitting even longer.
A logarithmic normalization (aka matplotlib's LogNorm) might be more useful here:
log_normalizer = mpl.colors.LogNorm(3, 90, clip=True)
log_values = log_normalizer(gdf[colname])
layer.get_fill_color = apply_continuous_cmap(log_values, BrBG_10)
layer.get_radius = log_values * 50
layer.radius_units = "meters"