Adding and changing columns · Pandas

Calculating from other columns

Python

import pandas as pd

orders = pd.DataFrame({
    "product": ["Keyboard", "Monitor", "Mouse", "Webcam"],
    "quantity": [1, 2, 3, 1],
    "price": [89.99, 279.00, 24.50, 59.00],
})
orders["total"] = orders["quantity"] * orders["price"]
orders["total_with_tax"] = (orders["total"] * 1.18).round(2)
print(orders)

Assigning to a new column name creates it, calculated for every row at once.

assign: several columns, as a new DataFrame

Python

import pandas as pd

orders = pd.DataFrame({
    "product": ["Keyboard", "Monitor", "Mouse", "Webcam"],
    "quantity": [1, 2, 3, 1],
    "price": [89.99, 279.00, 24.50, 59.00],
})
summary = orders.assign(
    total=lambda d: d["quantity"] * d["price"],
    share=lambda d: (d["total"] / d["total"].sum() * 100).round(1),
)
print(summary)

assign returns a new DataFrame and leaves the original alone, and later columns can use earlier ones, as share uses total. It suits long chains of steps.

Conditions: np.where

Python

import numpy as np
import pandas as pd

orders = pd.DataFrame({"product": ["Keyboard", "Monitor", "Mouse"], "total": [89.99, 558.00, 73.50]})
orders["size"] = np.where(orders["total"] >= 100, "large", "small")
orders["shipping"] = np.where(orders["total"] >= 100, 0, 49)
print(orders)

Lookups: map

Python

import pandas as pd

people = pd.DataFrame({"name": ["Asha", "Leo", "Mei", "Raj"], "city": ["Mumbai", "London", "Osaka", "Paris"]})
codes = {"Mumbai": "BOM", "London": "LHR", "Osaka": "KIX"}
people["airport"] = people["city"].map(codes)
print(people)

map with a dictionary replaces each value with its entry. Values with no entry, like Paris here, become missing.

Custom logic: apply

Python

import pandas as pd

orders = pd.DataFrame({
    "product": ["Keyboard", "Monitor", "Mouse"],
    "total": [89.99, 558.00, 73.50],
    "express": [True, False, True],
})

def delivery_fee(row):
    fee = 0 if row["total"] >= 100 else 49
    if row["express"]:
        fee += 99
    return fee

orders["fee"] = orders.apply(delivery_fee, axis=1)
print(orders)

apply(function, axis=1) calls your function once per row. It can express any logic, but it runs Python for every row, so on big tables prefer the column-at-a-time tools above.

Grouping numbers into bins: pd.cut

Python

import pandas as pd

students = pd.DataFrame({"name": ["Asha", "Leo", "Mei", "Raj", "Zoe"], "score": [92, 58, 74, 88, 75]})
students["grade"] = pd.cut(students["score"], bins=[0, 60, 75, 90, 100], labels=["F", "C", "B", "A"])
print(students)
print(students["grade"].value_counts().sort_index())

pd.cut sorts numbers into ranges. By default each bin includes its upper edge, so 75 falls in (60, 75] and gets a C.

Renaming, dropping and converting

Python

import pandas as pd

raw = pd.DataFrame({"Prod Name": ["Pen", "Bag"], "qty": ["3", "1"], "notes": ["", "gift"]})
clean = (raw
         .rename(columns={"Prod Name": "product", "qty": "quantity"})
         .drop(columns=["notes"])
         .astype({"quantity": int}))
print(clean)
print(clean["quantity"].sum())

rename takes a dictionary of old to new names, drop(columns=...) removes columns, and astype converts types. Chaining them, wrapped in brackets, reads as a sequence of cleaning steps.