Build a dashboard with Vega-Altair and Python

The Vega-Altair (opens in a new tab) is a great package for building interactive visualizations. This tutorial shows how to use it in Jupyter Notebook. Moreover, thanks to the Mercury framework, we will publish Python notebook as a web application.

Here is a preview of the notebook and the dashboard:

Import packages

We will need the following packages:

mercury
altair
vega_datasets

• mercury will be used to create interactive widgets and publish the notebook as a web app,
• altair is a plotting package,
• vega_datasets is a package with example datasets.

The import cell code:

import altair as alt
from vega_datasets import data
import mercury as mr

Setup app properties

In the next cell, we will setup App properties:

app = mr.App(title="Car report 📊", description="Report based on cars dataset")

The title and description will be used for displaying information about the app in the home view (when served with Mercury):

Load data

We will use the cars dataset from the vega_datasets package:

cars = data.cars()

Select samples based on widget selection

Let's add MultiSelect widget to choose origin of cars:

origin = mr.MultiSelect(label="Select origin", value=cars.Origin.unique(), choices=cars.Origin.unique())

We will use selection .value to filter the Data Frame:

df = cars[cars.Origin.isin(origin.value)]

Checkbox to control plots

We will add CheckBox to control number of plots:

show_bars = mr.Checkbox(label="Show bars plot", value=True)

Display metrics

Metrics about data can be displayed with NumberBox widget:

mr.NumberBox([
    mr.NumberBox(data=df.shape[0], title="Total cars"),
    mr.NumberBox(data=df.Horsepower.max(), title="Max Horse Power"),
    mr.NumberBox(data=df.Miles_per_Gallon.max(), title="Max MPG"),
])

Create Vega-Altair plots

Please use the following code to create plots about cars data:

points = alt.Chart(df).mark_point().encode(
    x='Horsepower:Q',
    y='Miles_per_Gallon:Q',
    color='Origin:N',
    tooltip=['Name', 'Origin', 'Horsepower', 'Miles_per_Gallon']
).interactive().properties(
    width="container"
)
 
bars = alt.Chart(df).mark_bar().encode(
    x='count()',
    y='Origin:N',
    color='Origin:N'
).properties(
    width="container"
)
 
charts = [points]
if show_bars.value:
    charts += [bars]
 
alt.vconcat(
    *charts
).configure(autosize="fit-x")

Display Data Frame samples

Please call the head() function to display the first 20 samples from our data:

df.head(20)

Here is the full code in the Jupyter Notebook:

Run as a Web App

Please start Mercury Server with the following command:

mercury run

You should see the Web App running at the address http://127.0.0.1:8000. Below is an example demo:

Deployment

There are several ways to deploy apps created in the Mercury. The simplest way is to use Mercury Cloud (opens in a new tab). It is a hosting service where you can deploy notebooks as web apps with a few clicks. The app from the tutorial is available at this address use-cases.runmercury.com (opens in a new tab).

There are several options for self-hosting Mercury. Please check documentation for details.