AM

Tutorials Blog Roll of Honour About

 

 

 

Labelling Line Charts in Python's Altair Package

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

The .mark_text() method doesn't work quite the other mark methods - .mark_circle(), .mark_bar(), etc - in altair. Sometimes it's quite simple, as in this example in the documentation. Other times it can be quite complex.

The Data

We're going to use the .iowa_electricity dataset, because it's very straightforward. there are three columns:

1. date, which is a date;
2. source, a categorical column with three values, and
3. net_generation, an integer value.

df = data.iowa_electricity()
df.head()

	year	source	net_generation
0	2001-01-01	Fossil Fuels	35361
1	2002-01-01	Fossil Fuels	35991
2	2003-01-01	Fossil Fuels	36234
3	2004-01-01	Fossil Fuels	36205
4	2005-01-01	Fossil Fuels	36883

c = alt.Chart(df).mark_line().encode(
    x='year:T',
    y='net_generation',
    color='source',
    tooltip=df.columns.to_list()).properties(
    title="Iowa Electricity",
    width=600,
    height=377)

c

Labelling the Lines

Modern charting theory suggests that it's a bit of a pain having to check colors against a legend in a line chart. Better to label the lines, as labels would then be found as the eye naturally tracks the line from left to right. But how to attach the labels? In the documentation example, labelling the data was easy because the chart was a bar chart. It's not so easy with a line chart. To correctly label these lines, we have to

1. Create a second chart to overlay on the first.
2. Attach the labels to this second chart.
3. Join them all together.

The Label Location Chart

holder = []
grouper = df.groupby('source')
for a, b in grouper:
    holder.append(b[b.year==b.year.max()])
    
df2 = pd.concat(holder)
df2

	year	source	net_generation
16	2017-01-01	Fossil Fuels	29329
33	2017-01-01	Nuclear Energy	5214
50	2017-01-01	Renewables	21933

c2 = alt.Chart(df2).mark_point().encode(
    x='year:T',
    y='net_generation')
c2

c + c2

c3 = c2.mark_text().encode(text='source:N')
c3

c + c2 + c3

Problems

1. Label alignment
2. Legend still exists
3. The points created by mark_point() aren't very pretty.

Label Alignment

c3 = c2.mark_text(dx=40).encode(text='source:N')
c3

A mistake I made in the video was not realising the Nuclear Energy label was closer to the left than the other other two labels. It's because mark_text() aligns to the middle by default, something I should have realised just by looking at the labels. But it's an easy fix with align='left'. And setting align='left' means we have to reduce our dx parameter too, as it's measuring against the leftmost part of the text now, rather than the center.

c3 = c2.mark_text(dx=10,
                 align='left').encode(text='source:N')
c3

c + c2 + c3

Eliminating the Point

As easy as setting size=0 as a parameter in .mark_point()

c2 = alt.Chart(df2).mark_point(size=0).encode(
    x='year:T',
    y='net_generation')
c + c2 + c3

And Finally - Eliminating the Legend

Just as the x parameter in .encode() can take an alt.X() method and the y an alt.Y(), so the color parameter can take a legend parameter. Here, we set color=alt.Color('source', legend=None), and our properly-labelled line chart is now complete.

c = alt.Chart(df).mark_line().encode(
    x='year:T',
    y='net_generation',
    color=alt.Color('source', legend=None),
    tooltip=df.columns.to_list()).properties(
    title="Iowa Electricity",
    width=600,
    height=377)

c + c2 + c3

The Original Video