Demo: IPyWidgets

The following sections contain examples of some common widget types. Remix this notebook to play around with them.

Note: IPyWidget support is currently experimental. Check out IPyWidgets Support for more info.

Interactive Matplotlib Visualisations

In its simplest form, widgets enhance Python functions with interaction capabilities. The following example shows how to use a context manager to plot solutions to a SIR ODE model (see hidden code cell) whose parameters can be changed by dragging the slider for each input.

SIR Model

Input Sliders

from ipywidgets import interact
@interact(beta=(0,1,0.05), gamma=(0,1,0.05), maxT=(5,100,5))
def plotSIR(beta = beta, gamma = gamma, maxT = maxT):
    '''Solve differential equations in SIR and plot'''
    t = np.linspace(0, maxT, 1000)
    soln = odeint(SIR, [S0,I0,R0], t, args=(beta, gamma))
    soln = np.array(soln)
    plt.plot(t, soln[:,0], linewidth=3, label = 'S(t)')
    plt.plot(t, soln[:,1], linewidth=3, label = 'I(t)')
    plt.plot(t, soln[:,2], linewidth=3, label = 'R(t)')
    plt.title("SIR model")
    plt.xlabel("Time"); plt.ylabel("proportions")
    plt.grid(); plt.legend()
IPyWidgets Demo (Python)

Compound Widgets

This examples shows how you can use a compound widget for exploring data frames. Specifically, we will use VBox and HBox for more fine-grained control over the layout and how to combine update callbacks with the observe method:

import pandas as pd
from ipywidgets import widgets
df = pd.read_csv(
IPyWidgets Demo (Python)

Data Frame Explorer

table = widgets.HTML()
html = widgets.HTML()
min_sl = widgets.IntSlider(min= 0, max=40, value=7)
max_sl = widgets.IntSlider(min=0, max=40, value=8)
def draw(_ = None):
  min_mpg = min_sl.value
  max_mpg = max_sl.value
  cols = ['ID', 'City mpg']
  selection = df.loc[df['City mpg'].between(min_mpg, max_mpg)]
  table.value = f'<h3>Found: {len(selection)}</h3>'
  html.value = selection[cols].to_html()
draw() # first draw 
              widgets.HBox([min_sl, max_sl]), 
IPyWidgets Demo (Python)

Plotly Figure Widgets

The following example shows how to set up a basic Plotly FigureWidget (based on the examples in Jon Mease's notebooks):

from sklearn import datasets
import time
iris_data = datasets.load_iris()
feature_names = [name.replace(' (cm)', '').replace(' ', '_') for name in iris_data.feature_names]
iris_df = pd.DataFrame(, columns=feature_names)
iris_class = + 1
iris dataframe setupIPyWidgets Demo (Python)

FigureWidgets behave almost identically to Plotly Figure objects. The major difference here is that they are also IPyWidgets that can be interacted with. If you execute all the cells following the next cell ("Commands ⌘/Ctrl+J → Run Cells Below") you’ll see the plot changing on each evaluation.

Mind you can always use "Commands ⌘/Ctrl+J → Outputs / Clear" to remove the plots.

import plotly.graph_objs as go
fig = go.FigureWidget()
Figure WidgetIPyWidgets Demo (Python)
fig.add_scatter(x=iris_df.sepal_length, y=iris_df.petal_width)
scatter =[0]
IPyWidgets Demo (Python)

Any change to the widget’s underlying data (which acts as traitlets) will automatically update the plot for you …

scatter.mode = 'markers'
scatter.marker.size = 8
IPyWidgets Demo (Python)
scatter.marker.color = iris_class
scatter.marker.cmin = 0.5
scatter.marker.cmax = 3.5
scatter.marker.colorscale = [[0, 'red'], [0.33, 'red'], 
                             [0.33, 'green'], [0.67, 'green'], 
                             [0.67, 'blue'], [1.0, 'blue']]
IPyWidgets Demo (Python)

… batch updates instead send a single message to the frontend, aggregating all changed attributes at once:

with fig.batch_update():
    scatter.marker.colorbar.ticks = 'outside'
    scatter.marker.colorbar.tickvals = [1, 2, 3]
    scatter.marker.colorbar.ticktext = iris_data.target_names.tolist()
    scatter.marker.colorbar.title = 'Species'
    scatter.marker.colorbar.titlefont.size = 16 = 'Rockwell'
    fig.layout.xaxis.title = 'sepal_length'
    fig.layout.yaxis.title = 'petal_width'
scatter.marker.showscale = True
IPyWidgets Demo (Python)

Animating Changes

All changes to the data can also be animated …

with fig.batch_animate(duration=1500):
    scatter.marker.size = np.sqrt(iris_df.petal_length.values * 60)
with fig.batch_animate(duration=1500):
    scatter.marker.size = 8
IPyWidgets Demo (Python)

… and Figure widgets can interact with any other kind of widget to be assembled into more complex visualisations:

from IPython.display import display, HTML
outx = widgets.Output()
outy = widgets.Output()
fig2 = go.FigureWidget(
IPyWidgets Demo (Python)
def report_extrema(trace, points, state):
    with outx:
      display(HTML(f"<em>Max x-value: {np.max(points.xs)}</em>"))
      display(HTML(f"<em>Min x-value: {np.min(points.xs)}</em>"))
    with outy:
      display(HTML(f"<em>Max y-value: {np.max(points.ys)}</em>"))
      display(HTML(f"<em>Min y-value: {np.min(points.ys)}</em>"))
fig2.layout.dragmode = 'lasso'[0].marker.showscale = False[0].on_selection(report_extrema)
IPyWidgets Demo (Python)

In the following plot, every "lasso" selection performed by the cursor will populate HTML output widgets with minima and maxima values from the selected points:

widgets.VBox([fig2, widgets.HBox([outx, outy])])
Lasso Selection CallbacksIPyWidgets Demo (Python)

We hope this short overview has been useful and inspires creating more interactive notebooks.


If widget display goes out-of-sync with the backend kernel try to:

  • Clear outputs: Commands ⌘/Ctrl+J → Outputs / Clear

  • Stop the runtime: Commands ⌘/Ctrl+J → Runtime / Reset


Runtimes (1)