AM

Tutorials Blog Roll of Honour About

 

 

 

Pie Charts and Encodings in Python's Altair Package

Standard Imports

import pandas as pd
import altair as alt
from vega_datasets import data

Altair Version

As pie charts are a relatively recent addition to Altair - the standing of pie charts is not high among in charting/plotting academia - we need to check out Altair version to make sure it's up to date. If it isn't, updating is easy - just open a terminal and run pip install -U altair.

alt.__version__

'4.2.0'

Seattle Weather Data

The Seattle weather data has six fields. To these, we're going to add a sixth, year, as we're going to use that year category to slice our pie.

Note that we're converting out year field to int from str. That'll be important later.

df = data.seattle_weather()
df['year'] = df.date.apply(lambda x: int(x.strftime('%Y')))
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1461 entries, 0 to 1460
Data columns (total 7 columns):
 #   Column         Non-Null Count  Dtype         
---  ------         --------------  -----         
 0   date           1461 non-null   datetime64[ns]
 1   precipitation  1461 non-null   float64       
 2   temp_max       1461 non-null   float64       
 3   temp_min       1461 non-null   float64       
 4   wind           1461 non-null   float64       
 5   weather        1461 non-null   object        
 6   year           1461 non-null   int64         
dtypes: datetime64[ns](1), float64(4), int64(1), object(1)
memory usage: 80.0+ KB

Creating the Pie Chart

While the other .mark_*() methods we've seen so far all use x and y parameters, .mark_arc() doesn't. It uses:

• theta, which is set to the numerical field that will be shown the chart, and
• color, which is set to the categorical field by which we'll slice the chart.

It's really a very simple chart.

c = alt.Chart(df).mark_arc().encode(
    theta='sum(precipitation)',
    color='year')
c

A Question of Encoding

This pie chart isn't quite what you were expecting - or at least, it shouldn't be. Altair as recognised year as a number. It's thousand-separated the digits - "2,015" rather than "2015" - and it's used a color scale instead of discrete colors. This isn't what we want. We want it to see year as a discrete category. There are two ways to do this.

In the Data Frame

The first way is to change the data dtype(dtype) in the data frame itself - or, in this particular case, to have left the year as a string rather than converting it to an integer when we called our lambda function. And that's fine, and often the best way to do it.

Using Altair's Own Encoding Data Types

Altair has a shorthand that covers five data types, and allows users to explicitly state if they'd like a particular category treated in a particular way by suffixing a colon and one of five letters to the column name in setting the parameter = x=something:O, y=something:N, z=something:Q, or whather. These are the five abbreviations.

Abbreviation	Meaning	Example
N	Nominal	These are the names of things - any thing at all. Cats, dogs, apples, oranges.
O	Ordinal	These are the names of things that we associate as having a certain order. Days of the week, months of the year, gold-silver-bronze, those sorts of things.
Q	Qualitative	Qualitative data is any numerical data at all.
T	Temporal	Temporal data is datetime data.
G	geographical	Geographical Data is .geojson data, latitudes and longitudes.

And here's the pie chart proper, with color set to year:N.

c = alt.Chart(df).mark_arc().encode(
    theta='sum(precipitation)',
    color='year:N',
tooltip=['year', 'sum(precipitation)'])
c

The Original Video