Bar Charts for Categories

A bar chart is useful when you want to compare separate categories. In the study log, subject is a category:

math, history, programming, writing

The question:

Which subjects have the most total study time?

is a category comparison. This chapter uses a small sample log:

import pandas as pd

records = [
    {"subject": "math", "minutes": 45, "date": "2026-05-10"},
    {"subject": "history", "minutes": 30, "date": "2026-05-11"},
    {"subject": "programming", "minutes": 60, "date": "2026-05-11"},
    {"subject": "math", "minutes": 25, "date": "2026-05-12"},
    {"subject": "programming", "minutes": 90, "date": "2026-05-13"},
    {"subject": "writing", "minutes": 45, "date": "2026-05-14"},
]

df = pd.DataFrame(records)

First, summarize the DataFrame:

subject_totals = df.groupby("subject")["minutes"].sum()
subject_totals = subject_totals.sort_values(ascending=False)
print(subject_totals)

Then plot the summary:

import matplotlib.pyplot as plt

fig, ax = plt.subplots()
ax.bar(subject_totals.index, subject_totals.values)
ax.set_title("Total study minutes by subject")
ax.set_xlabel("Subject")
ax.set_ylabel("Minutes")
plt.show()

The chart is not plotting every session. It is plotting one summary value per subject:

From Summary Rows to Bars

It is worth tracing the connection once by hand. The sorted summary prints as:

subject
programming    150
math            70
writing         45
history         30
Name: minutes, dtype: int64

The chart draws exactly these rows, in this order:

• the first row, programming 150, becomes the first bar, 150 units tall,
• the second row, math 70, becomes the second bar,
• the third row, writing 45, becomes the third bar,
• and the last row, history 30, becomes the shortest bar.

Nothing else enters the chart. If a bar looks wrong, the summary row behind it is wrong, and the place to look is the groupby step, not the plotting call. This is the same layer-by-layer debugging you used with pandas summaries in Using Libraries.

Why Sorting Helps

Compare these two printed summaries:

subject
history         30
math            70
programming    150
writing         45
Name: minutes, dtype: int64

and:

subject
programming    150
math            70
writing         45
history         30
Name: minutes, dtype: int64

The second one is easier to scan because the biggest value is first. Sorting helps the chart for the same reason.

subject_totals = subject_totals.sort_values(ascending=False)

Do this before plotting. The bar order comes from the order of subject_totals.index.

Here is the same chart without sorting. Every number is identical, but the reader has to work harder to compare the subjects:

A Helper for Plotting Data

It is often useful to prepare a small DataFrame for the chart:

summary = (
    df.groupby("subject", as_index=False)["minutes"]
    .sum()
    .sort_values("minutes", ascending=False)
)

This produces a table with two columns:

       subject  minutes
2  programming      150
1         math       70
3      writing       45
0      history       30

The as_index=False part keeps subject as an ordinary column instead of making it the Series index. That can be convenient when using seaborn later, because seaborn usually names columns directly.

summary = df.groupby("subject", as_index=False)["minutes"].sum()

Now apply the same preparation steps to library-loan data. The column names and subject matter change, but the grouped-and-sorted result still has one row per category.

df = pd.DataFrame([
    {"genre": "mystery", "loans": 12, "date": "2026-05-10"},
    {"genre": "history", "loans": 8, "date": "2026-05-11"},
    {"genre": "mystery", "loans": 7, "date": "2026-05-12"},
])

     genre  loans
0  mystery     19
1  history      8

Common Mistake: Plotting Raw Rows

It is tempting to skip the summary and plot the DataFrame directly:

These are the six rows that would be passed to the plotting call:

subject      minutes
math              45
history           30
programming       60
math              25
programming       90
writing           45

fig, ax = plt.subplots()
ax.bar(df["subject"], df["minutes"])
ax.set_title("Raw study-session rows")
ax.set_xlabel("Subject")
ax.set_ylabel("Minutes")
plt.show()

Here is what those raw rows produce. Repeated subjects occupy the same x position, so some bars cover others:

This runs without an error, but matplotlib draws six rectangles at only four subject positions. Rows that share a subject overlap. Matplotlib does not aggregate the rows or calculate a maximum. Because these same-color rectangles overlap, the visible outline reaches the tallest individual rectangle at each subject position. Here, math appears to be 45 minutes instead of its 70-minute total, and programming appears to be 90 minutes instead of its 150-minute total. The result looks like a subject comparison while not answering the question “which subject has the most minutes?”

If the question is about totals, group first, then plot the summary. When a bar chart surprises you, check what one bar actually represents before adjusting any plotting code.

ax.bar(df["subject"], df["minutes"])

What Bar Charts Are Not For

A bar chart is not always the best first choice.

If the question is about time order:

How did study minutes change over the week?

a line chart is usually clearer than a bar chart of subjects.

If the question is about individual session lengths:

Are sessions usually short or long?

a histogram is usually clearer than a bar chart with one bar per session.

If the categories are not comparable, a bar chart can mislead. For example, comparing total minutes for “math” and “all science subjects” would mix one subject with a whole group. The bars would look comparable, but the categories would not mean the same kind of thing.

Before making a bar chart, check:

1. Is each bar a comparable category?
2. Does each bar measure the same value, such as total minutes?
3. Would sorting make the comparison easier?
4. Are the labels clear enough that the reader knows what is being compared?

If yes, a bar chart is a good tool.

Check Your Understanding

1. Does a subject-total bar chart plot raw sessions or a grouped summary?
2. Why does sorting the totals before plotting help the reader?
3. What does as_index=False change about the grouped result?
4. Name a study-log question where a bar chart is the wrong first choice.