ContextQMD
Back to Deck Gl

Selecting data for use in Jupyter: Singapore taxi location data

bindings/pydeck/examples/03 - Event handlers and data selection in pydeck.ipynb

9.3.22.8 KB
Original Source

Selecting data for use in Jupyter: Singapore taxi location data

Often it's easier to use a visual application to draw a shape than define a geometry in code. Here we'll explore how pydeck can be used to select data and pass that selected data back to the Jupyter kernel for use in Pandas.

Currently this notebook runs against live data, and there is a slight chance that the source data will be unavailable.

Contents

Getting the data

Here we'll use the live taxi location API provided by the government of Singapore.

python
import pydeck as pdk

DATA_URL = 'https://api.data.gov.sg/v1/transport/taxi-availability'
COLOR_RANGE = [
  [255, 255, 178, 25],
  [254, 217, 118, 85],
  [254, 178, 76, 127],
  [253, 141, 60, 170],
  [240, 59, 32, 212],
  [189, 0, 38, 255]
]
python
import pandas as pd
import requests

json = requests.get(DATA_URL).json()
df = pd.DataFrame(json["features"][0]["geometry"]["coordinates"])
df.columns = ['lng', 'lat']

viewport = pdk.data_utils.compute_view(df[['lng', 'lat']])
layer = pdk.Layer(
    'ScreenGridLayer',
    df,
    get_position=['lng', 'lat'],
    cell_size_pixels=20,
    color_range=COLOR_RANGE,
    pickable=True,
    auto_highlight=True)
r = pdk.Deck(layers=[layer], initial_view_state=viewport)
python
r.show()

Two way communication

Click the above visualization and then execute the cell below to pass data from the application to Python.

python
r.selected_data

Event handlers

You can also create more complex applications by drawing from the many widgets available within the ipywidgets library. Currently, pydeck supports a number of event handlers which can be linked to ipywidgets or elsewhere.

Here, we'll use ipywidgets to tell us the number of taxis in the current viewport.

python
from ipywidgets import HTML

text = HTML(value='Points in viewport:')
layer = pdk.Layer('ScatterplotLayer', df, get_position=['lng', 'lat'], get_fill_color=[255, 0, 0], get_radius=100)
r = pdk.Deck(layer, initial_view_state=viewport)
python
def filter_by_bbox(row, west_lng, east_lng, north_lat, south_lat):
    return west_lng < row['lng'] < east_lng and south_lat < row['lat'] < north_lat


def filter_by_viewport(widget_instance, payload):
    try:
        west_lng, north_lat = payload['data']['nw']
        east_lng, south_lat = payload['data']['se']
        filtered_df = df[df.apply(lambda row: filter_by_bbox(row, west_lng, east_lng, north_lat, south_lat), axis=1)]
        text.value = 'Points in viewport: %s' % int(filtered_df.count()['lng'])
    except Exception as e:
        text.value = 'Error: %s' % e


r.deck_widget.on_view_state_change(filter_by_viewport)
display(text)
r.show()