Grouping

Split, apply, combine

groupby is the single most useful thing in pandas. It splits the rows into groups, applies an aggregation to each, and combines the answers:

df.groupby("country")["sales"].sum()
df.groupby("country")["sales"].mean()
df.groupby("country").size()             # rows per group

Group by several columns for a finer breakdown, and use .agg when one aggregation is not enough:

df.groupby(["country", "year"])["sales"].sum()
df.groupby("country")["sales"].agg(["sum", "mean", "count"])

The result is indexed by the grouping key. .reset_index() turns that index back into an ordinary column, which is usually what you want before printing or exporting.

Your turn: total the sales per region, and find which region sold most.

Several aggregations at once

.agg takes a dictionary mapping each column to what you want from it:

df.groupby("region").agg({"sales": "sum", "units": "mean"})

Your turn: write summarise(df) returning, per region, the total sales and the number of rows, as a dictionary of the form {"north": {"sales": 335, "orders": 3}, ...}.