Lesson 11

Bar Charts

Compare separate categories with plt.bar, control colour and width, and understand why a bar chart must always start at zero.

Intermediate25 min

A bar chart compares quantities across categories that have no natural order: teams, products, regions, tracks. Each bar's length is the number, and because the eye compares lengths well, a bar chart is usually the clearest answer to which is bigger, and by how much.

A bar chart of learners enrolled on four tracks

Categories and heights

plt.bar(x, height) takes the category labels and the numbers. Matplotlib spaces the bars evenly and writes the labels under them, so text categories need no extra work.

python

import matplotlib.pyplot as plt

tracks = ["Python", "NumPy", "pandas", "Plotting"]
enrolled = [120, 86, 94, 58]

plt.bar(tracks, enrolled)
plt.title("Learners per track")
plt.xlabel("track")
plt.ylabel("learners")
plt.show()

print("largest:", tracks[enrolled.index(max(enrolled))])
print("total:", sum(enrolled))

Output:

text

largest: Python
total: 358

Colour, one or many

One colour string paints every bar. A list of colours paints them one by one, which is worth doing only when the colour means something, such as highlighting the bar you are writing about. A rainbow of colours that carry no meaning just slows a reader down.

python

import matplotlib.pyplot as plt

tracks = ["Python", "NumPy", "pandas", "Plotting"]
enrolled = [120, 86, 94, 58]

plt.bar(tracks, enrolled, color="#2563EB")
plt.title("One colour: the lengths do the talking")
plt.show()

quiet = "#CBD5E1"
shades = [quiet if track != "pandas" else "#7C3AED" for track in tracks]
plt.bar(tracks, enrolled, color=shades)
plt.title("Colour used to point at one bar")
plt.show()

print("highlighted:", "pandas")

Output:

text

highlighted: pandas

A bar chart with the third bar highlighted and the rest grey

Width and gaps

width is measured in category slots, and the default is 0.8, which leaves a small gap between bars. Narrower bars give more white space; a width of 1 removes the gaps entirely and starts to look like a histogram, which is a different chart with a different meaning.

python

import matplotlib.pyplot as plt

tracks = ["Python", "NumPy", "pandas", "Plotting"]
enrolled = [120, 86, 94, 58]

plt.bar(tracks, enrolled, width=0.5, color="#7C3AED")
plt.title("width=0.5")
plt.show()

print("default width:", 0.8)
print("gaps say the categories are separate things")

Output:

text

default width: 0.8
gaps say the categories are separate things

Sorting and a fair zero

Two habits make bar charts honest and easy to read. Sort the bars by size unless the categories have their own order, so the ranking is visible. And leave the value axis starting at zero, because a bar that starts at 80 is telling the reader its length means something it does not.

python

import matplotlib.pyplot as plt

tracks = ["Python", "NumPy", "pandas", "Plotting"]
enrolled = [120, 86, 94, 58]

ordered = sorted(zip(enrolled, tracks), reverse=True)
heights = [value for value, _ in ordered]
labels = [name for _, name in ordered]

plt.bar(labels, heights, color="#7C3AED")
plt.title("Sorted, and starting at zero")
plt.ylabel("learners")
plt.grid(axis="y", alpha=0.3)
plt.show()

for name, value in zip(labels, heights):
    print(f"{name:<9} {value}")

Output:

text

Python    120
pandas    94
NumPy     86
Plotting  58

Key takeaways

plt.bar(x, height) compares categories that have no natural order
One colour string paints every bar; a list paints them individually
Use colour to point at a bar, not to decorate every one
Sort by size, and always start the value axis at zero
Next: Horizontal Bar Charts
Previous: Working with Figure and Axes Objects

Loading playground…

Check your understanding

Bar Charts Quiz

What is a bar chart for?

1 / 10

← Working with Figure and Axes Objects

Horizontal Bar Charts →