PyPath
Start learning
Unit 4 • Lesson 4

Adding and Dropping Columns

Overview

A new column is an assignment, and drop returns a new table by default.

Intermediate 20–25 min
← Previous Next: dtypes in pandas

What You Will Learn in This Lesson

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

  • Create a derived column from the columns you already have.
  • Say what drop returns, and what it leaves behind.
  • Take a copy before modifying a table a caller still owns.
  • Rename a column without rebuilding the table.

Why This Matters

The column that answers your question is usually not in the file — it is a rate, a difference, a share. Deriving one is a single assignment, and the only real trap is that half of these operations return a new table while the other half change yours, and the difference is not visible until something downstream is wrong.

Step 1: A derived column

Assigning to a name that does not exist creates it. The right-hand side is ordinary column arithmetic, applied to the whole column at once.

1

Create by assigning

No method call and no declaration. If the name exists it is replaced; if not, it is added at the end.

import pandas as pd

df = pd.DataFrame({"price": [10.0, 20.0]})
df["with_tax"] = df["price"] * 1.08
print(df)
2

From two columns

The arithmetic aligns on the index, so the rows always line up with themselves. This is the everyday case: a total, a rate, a difference from a target.

import pandas as pd

df = pd.DataFrame({"price": [2.0, 5.0], "quantity": [3, 2]})
df["total"] = df["price"] * df["quantity"]
print(df)
3

Renaming

rename takes a mapping and, like drop, hands back a new table unless you tell it otherwise.

import pandas as pd

df = pd.DataFrame({"n": [1, 2]})
print(df.rename(columns={"n": "count"}).columns.tolist())
print(df.columns.tolist())   # unchanged

Mini Practice #1: A rate per head

Try It Yourself

Add a per_person column equal to total divided by people, then a rounded column holding it to one decimal place.

Press Run to see output

Step 2: Dropping returns a copy

drop hands back a new table and leaves yours alone unless you reassign. This surprises people who expect it to behave like a list method.

1

The result is the point

Calling drop and ignoring what it returns looks exactly like a no-op, because it is one.

import pandas as pd

df = pd.DataFrame({"a": [1], "b": [2]})
smaller = df.drop(columns=["b"])
print(list(smaller.columns))
print(list(df.columns))      # still both
2

Assignment is not a copy

Two names for the same table are still one table. A function that assigns a column onto the DataFrame it was handed has changed its caller's data, and nothing says so.

import pandas as pd

def add_flag(table):
    table["flag"] = True        # edits the caller's table
    return table

df = pd.DataFrame({"a": [1]})
add_flag(df)
print(list(df.columns))
3

Copy first

One extra line makes the function honest: it returns something new and leaves what it was given alone.

import pandas as pd

def add_flag(table):
    out = table.copy()
    out["flag"] = True
    return out

df = pd.DataFrame({"a": [1]})
add_flag(df)
print(list(df.columns))      # untouched

Mini Practice #2: Which one changed

Try It Yourself

Run this and note which of the two prints shows the extra column. Then add .copy() and run it again.

Press Run to see output

Step 3: When to Use It

A derived column

Assign an expression over existing columns to a new name.

df["percent"] = df["score"] / df["total"] * 100

Removing columns you will not use

drop(columns=[...]) returns a smaller table that is easier to read and check.

When Not to Use It

Do not expect drop to change the table in place. It returns a new DataFrame — assign the result, or the column is still there.

End-of-Lesson Exercises

Exercise 1: Add a total, leave the original alone

Write add_total(df) that returns a new table with a total column equal to price times quantity, leaving the table it was given unchanged.

Take a copy first, then assign the new column onto the copy. Assigning straight onto df would change the caller's table, which the hidden case checks.

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

Exercise 2: Keep only what you need

Write only(df, keep) that returns a new table holding just the columns named in keep, in that order, without changing the table it was given.

df[keep] with a list of names already gives a new table in the order you asked for. Add .copy() so later edits to the result cannot reach back.

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