Skip to content

Get Started

What is Jupyter Scatter?

Jupyter Scatter is a scalable, interactive, and interlinked scatter plot widget exploring datasets with up to several million data points that runs in Jupyter Lab/Notebook and Google Colab. It focuses on data-driven visual encodings and offers two-way pan+zoom and lasso interactions. Beyond a single plot, Jupyter Scatter can compose multiple scatter plots and synchronize their views and point selections.

Key Features

  • 🖱️ Interactive: Pan, zoom, and select data points interactively.
  • 🚀 Scalable: Plot up to several millions data points smoothly.
  • 🔗 Interlinked: Synchronize the view, hover, and selection across multiple plots.
  • Effective Defaults: Perceptually effective point colors and opacity by default.
  • 📚 Friendly API: A readable API that integrates deeply with Pandas DataFrames.
  • 🛠️ Integratable: Use Jupyter Scatter in your own widgets by observing its traitlets.

Simplest Example

In the simplest case, you can pass the x/y coordinates to the plot function as follows:

import jscatter
import numpy as np

x = np.random.rand(500)
y = np.random.rand(500)

jscatter.plot(x, y)

Bind a Pandas DataFrame

In most cases, however, it's more convenient to work with a DataFrame and reference the x/y columns via their names.

import jscatter
import numpy as np
import pandas as pd

df = pd.DataFrame(np.random.rand(500, 2), columns=['mass', 'speed'])

jscatter.plot(data=df, x='mass', y='speed')

Point Color, Opacity, and Size

Often times, we want to style the points. Jupyter Scatter allows you to do this as follows:

    color='red', # static visual encoding
    size=10, # static visual encoding
    opacity=0.5 # static visual encoding

However, more commonely, one wants to use these three point attributes (color, opacity, and size) to visualize data properties.

For instance, in the following we're extending the data frame with a continuous property called pval and a categorical property called cat.

df = pd.DataFrame({
  # Random floats
  "mass": np.random.rand(500),
  "speed": np.random.rand(500),
  "pval": np.random.rand(500),
  # Random letters A, B, C, D, E, F, G, H
  "cat": np.vectorize(lambda x: chr(65 + round(x * 7)))(np.random.rand(500)),

You can visualize the two properties by referencing their columns using the color_by, opacity_by, or size_by arguments.

    color_by='cat', # data-driven visual encoding
    size_by='pval', # data-driven visual encoding

Notice how jscatter uses a reasonable color and size map by default. Both are based on the properties' data types. In this examples, the jscatter picked the color blindness safe color map from Okabe and Ito as the number of categories is less than 9.

When visualizing the pval via the color we see how the default color map switches to Viridis given that pval is a continuous property.

    color_by='pval', # pval is continuous data
    size_by='pval', # pval is categorical data

You can of course customize the color map and many other parameters of the visual encoding as shown next.

    # Custom categorical color map 
      A='red',    B='#00ff00', C=(0,0,1),   D='DeepSkyBlue',
      E='orange', F='#702AF7', G='#2AF7C0', H='teal'
    # Custom size map (specified as a linspace)
    size_map=(2, 20, 10),

Functional API

The flat API (that we used before) can get overwhelming when we customize a lot of properties. Therefore, jscatter provides a functional API that groups properties by type and exposes them via meaningfully-named methods that can almost be read like a sentence.

For instance, in line two of the example below, the scatter plot colors points by the mass column by mapping its values to the plasma color map in reverse order.

scatter = jscatter.Scatter(data=df, x='mass', y='speed')
scatter.color(by='mass', map='plasma', order='reverse')
scatter.size(by='pval', map=[2, 4, 6, 8, 10])

Update Properties After Plotting

You don't have to specify all properties upfront. Using the functional API you can update scatter plot instances after having plotted the scatter and the plot will automatically re-render.

For instance, in the following we're changing the color map to magma in reverse order.

scatter.color(map='magma', order='reverse')

Chaining Method Calls

Inspired by D3 you can also chain methods calls as follows to update multiple property groups at once.


Animating Point Coordinates

When you update the x/y coordinates dynamically and the number of points match, the points will animate in a smooth transition from the previous to their new point location.

For instance, try calling scatter.xy('speed', 'mass') and you will see how the points are mirrored along the diagonal.

Retrieving Properties

Lastly, all method arguments are optional. If you specify arguments, the methods will act as setters and change the properties. However, if you call a method without any arguments it will act as a getter and return the related properties.

For example, scatter.color() will return the current coloring settings.

{'default': (0, 0, 0, 0.66),
 'selected': (0, 0.55, 1, 1),
 'hover': (0, 0, 0, 1),
 'by': 'mass',
 'map': [[0.001462, 0.000466, 0.013866, 1.0],
  [0.002258, 0.001295, 0.018331, 1.0],
  [0.987387, 0.984288, 0.742002, 1.0],
  [0.987053, 0.991438, 0.749504, 1.0]],
 'norm': <matplotlib.colors.Normalize at 0x15f23feb0>,
 'order': 'reverse',
 'labeling': None}