Easy Plotly

Build Status codecov.io Language grade: Python Pypi pyversions Jupyter Notebook GitHub.io Star

This is on-going research on how ploting with Plotly.py, especially ploting of hierarchical data, could be made easier.

See the outputs of the commands below - tables and plots - in the HTML export of this notebook. Or even, open this README.md as a notebook and run it interactively on Binder!


Install the easyplotly python package with

pip install easyplotly

Sample data

Our sample data is the population and life expectancy, per country and region:

In [1]:
import world_bank_data as wb
import itables.interactive

# Collect countries
countries = wb.get_countries()
region_country = countries[['region', 'name']].rename(columns={'name': 'country'})

# Population & life expectancy
region_country['population'] = wb.get_series('SP.POP.TOTL', mrv=1, id_or_value='id', simplify_index=True)
region_country['life_expectancy'] = wb.get_series('SP.DYN.LE00.IN', mrv=1, id_or_value='id', simplify_index=True)

# Observations restricted to the countries
pop_and_exp = region_country.loc[countries.region != 'Aggregates'].set_index(['region', 'country']).sort_index()
population life_expectancy
region country

Sunburst Charts

In [2]:
import plotly.graph_objects as go
import plotly.io as pio
import easyplotly as ep

pio.renderers.default = 'notebook_connected'
layout = go.Layout(title='World Population and Life Expectancy<br>Data from the World Bank', height=800)

Our Sunburst function accepts inputs of many types: pandas Series, dictionaries, and list of such objects. If you want, you can redefine labels, or add other arguments like text - use either a Series with an index identical to that of values, or a function that to any tuple (level0, level1, ... leveln) associates the corresponding label or value.

In [3]:
sunburst = ep.Sunburst(pop_and_exp.population, text=pop_and_exp.life_expectancy)
go.Figure(sunburst, layout)


The Treemap function works like the Sunburst one:

In [4]:
treemap = ep.Treemap(pop_and_exp.population, text=pop_and_exp.life_expectancy)
go.Figure(treemap, layout)

Just like the Sunburst function, it also accepts all the arguments supported by the original go.Sunburst object. You're even welcome to use the magic underscore notation, as we do below when we set marker.colors with marker_colors:

In [5]:
import numpy as np

def average(values, weights):
    """Same as np.average, but remove nans"""
    total_obs = 0.
    total_weight = 0.
    if isinstance(values, np.float):
        values = [values]
        weights = [weights]
    for x, w in zip(values, weights):
        xw = x * w
        if np.isnan(xw):
        total_obs += xw
        total_weight += w
    return total_obs / total_weight if total_weight != 0 else np.NaN

def life_expectancy(item):
    """Life expectancy associated to a tuple like (), ('Europe & Central Asia') or ('East Asia & Pacific', 'China')"""
    sub = pop_and_exp.loc[item] if item else pop_and_exp
    return average(sub.life_expectancy, weights=sub.population)

def text(item):
    """Return the text associated to a tuple like (), ('Europe & Central Asia') or ('East Asia & Pacific', 'China')"""
    life_exp = life_expectancy(item)
    if life_exp > 0:
        pop = pop_and_exp.population.loc[item].sum() if item else pop_and_exp.population.sum()  
        return 'Population: {:,}<br>Life expectancy: {:.2f}'.format(int(pop), life_exp)

treemap = ep.Treemap(pop_and_exp.population,
                     # magic underscore notation

go.Figure(treemap, layout)

Treemaps and Sunburst also accept trees with a non-constant depth - use a dictionary indexed with tuples of varying size. For instance, here is a tree that represents the files in this project:

In [6]:
import os
from plotly.colors import DEFAULT_PLOTLY_COLORS

project_files = os.popen('git ls-tree --name-only -r master').read().split()

size = {tuple(path.split('/')):os.stat(path).st_size for path in project_files}
log_size = {i: np.log(size[i]) for i in size}
extensions = set(os.path.splitext(path)[1] for path in project_files)
color_map = dict(zip(extensions, DEFAULT_PLOTLY_COLORS))

def node_color(node):
    if not node:
    node_extension = os.path.splitext(node[-1])[1]
    return color_map[node_extension]
In [7]:
treemap = ep.Treemap(log_size,


Sankey Plot

Plot links from a dict, or a series with a source/target multiindex:

In [8]:
links = {('A', 'B'): 3, ('B', 'C'): 1, ('B', 'D'): 2, ('C', 'A'): 1, ('D', 'A'): 1, ('A', 'D'): 1}

Plot links from a DataFrame (sources as the index, targets as the columns):

In [9]:
import pandas as pd
In [10]:
links = pd.DataFrame(1, index=['Source A', 'Source B'], columns=['Target'])

We conclude the examples with a plot in which the links are a list of pandas Series:

In [11]:
region_income = wb.get_countries().query("region != 'Aggregates'").copy()
region_income['population'] = wb.get_series('SP.POP.TOTL', mrv=1, id_or_value='id', simplify_index=True)
income_lending = region_income.copy()
region_income.set_index(['region', 'incomeLevel'], inplace=True)
income_lending.set_index(['incomeLevel', 'lendingType'], inplace=True)

layout = go.Layout(title='Regions income and lending type<br>Data from the World Bank')

sankey = ep.Sankey(
    link_value=[region_income['population'], income_lending['population']],
    link_label=[region_income['name'], income_lending['name']])

go.Figure(sankey, layout)