PyPath
Start learning
Unit 9 • Lesson 4

Resampling

Overview

Group by time period without building the key yourself.

Intermediate 20–25 min
← Previous Next: Rolling Windows

What You Will Learn in This Lesson

By the end of this lesson, you will be able to:

  • Bucket rows into time periods.
  • Say why a period with no rows still appears.
  • Pick the right frequency string.
  • Explain why that matters for a chart.

Why This Matters

A group-by on a month key omits months with no data. Plotted, that gap becomes a straight line between two points that were never adjacent, and the chart shows a trend that did not happen. resample produces every period in the range, so a quiet month is visible as a zero rather than invisible.

Step 1: resample is group-by for time

With a datetime index, resample buckets rows into periods — monthly, weekly, daily — and then you name the summary, exactly as with a group-by.

1

The call

A frequency string and a summary. "MS" is month start, "W" weekly, "D" daily, "YE" year end.

import pandas as pd

idx = pd.to_datetime(["2024-01-05", "2024-01-20", "2024-02-01"])
s = pd.Series([1, 2, 5], index=idx)
print(s.resample("MS").sum())
2

Any summary, not just sum

Mean, max, count — whatever you would ask a group for.

import pandas as pd

idx = pd.to_datetime(["2024-01-05", "2024-01-20", "2024-02-01"])
s = pd.Series([1, 2, 5], index=idx)
print(s.resample("MS").agg(["sum", "mean", "size"]))

Mini Practice #1: Monthly and weekly

Try It Yourself

Resample this series by month and then by week, and compare how many buckets each gives.

Press Run to see output

Step 2: Empty periods appear

Unlike a group-by, resample produces every period in the range, including the ones with no rows.

1

The quiet month

January and March have data; February does not. resample reports it as 0; a group-by would leave it out entirely.

import pandas as pd

idx = pd.to_datetime(["2024-01-01", "2024-03-01"])
s = pd.Series([1, 1], index=idx)

print(s.resample("MS").sum())
print(s.groupby(s.index.month).sum())

Two points on a chart from the group-by, three from the resample. The two-point version draws a line straight through a month that had nothing in it.

2

Zero is not always the honest fill

For a count, an empty period really is zero. For a mean it is unknown, and sum() quietly turning it into 0 would claim a month averaged nothing.

import pandas as pd

idx = pd.to_datetime(["2024-01-01", "2024-03-01"])
s = pd.Series([10.0, 30.0], index=idx)
print(s.resample("MS").sum())    # February reads 0
print(s.resample("MS").mean())   # February reads NaN, which is true

Mini Practice #2: The missing month

Try It Yourself

Compare the resample with the group-by and count the buckets each produces.

Press Run to see output

Step 3: When to Use It

Totals per period

Daily rows to weekly or monthly totals without building a period column.

df["sales"].resample("MS").sum()

Finding gaps

resample includes periods with no rows, so size() shows where data is missing.

When Not to Use It

Do not read a zero in a resampled sum as “nothing happened”. An empty period also sums to zero — check size() to tell them apart.

End-of-Lesson Exercises

Exercise 1: Monthly totals, gaps included

Write monthly_totals(s) that takes a Series indexed by date and returns a list of monthly totals, including months with nothing in them.

resample("MS") buckets by month start. The quiet month in the middle is the whole point: a group-by would skip it and the timeline would lie.

Press Run to try it, then Check my work to mark it.

Exercise 2: How many periods were empty

Write quiet_months(s) returning how many months in the range had no rows at all. Use a summary where an empty period is distinguishable from a period that summed to zero.

resample("MS").size() gives the row count per period, and the empty ones are the zeros. Using sum() instead would confuse a quiet month with one whose values happened to cancel out.

Press Run to try it, then Check my work to mark it.
← Prev Next