Pandas Fundamentals in 40 Minutes

import pandas as pd

# 1. Load
df = pd.read_csv('data.csv')

# 2. Peek
df.head()
df.info()
df.describe()

# 3. Select columns
df['age']                      # one column (Series)
df[['age', 'income']]          # multiple columns (DataFrame)

# 4. Filter rows
df[df['age'] > 18]
df[(df['age'] > 18) & (df['country'] == 'US')]

# 5. Sort
df.sort_values('income', ascending=False)

# 6. Create columns
df['income_per_age'] = df['income'] / df['age']

# 7. Group and aggregate
df.groupby('country')['income'].mean()
df.groupby(['country', 'gender']).agg({
    'income': ['mean', 'median'],
    'age': 'mean'
})

# 8. Join tables
merged = pd.merge(df, other_df, on='user_id', how='left')

# 9. Pivot
pd.pivot_table(df, index='country', columns='year', values='income')

# 10. Save
df.to_csv('clean.csv', index=False)
df.to_parquet('clean.parquet')

# .loc uses labels
df.loc[5]                      # row with index label 5
df.loc[df['age'] > 18, 'name'] # name column, filtered rows

# .iloc uses positions
df.iloc[5]                     # 6th row regardless of index label
df.iloc[:10, :3]               # first 10 rows, first 3 cols

# Chained assignment is a trap
# df[df.age > 18]['score'] = 100  # DO NOT DO THIS
df.loc[df.age > 18, 'score'] = 100  # CORRECT

# Top N per group
top3 = df.groupby('country').apply(
    lambda g: g.nlargest(3, 'income')
).reset_index(drop=True)

# Rolling stats
df['7d_avg'] = df['sales'].rolling(window=7).mean()

# Replace based on mapping
df['country'] = df['country'].replace({'USA': 'US', 'U.S.A.': 'US'})

# One-hot encoding
df_encoded = pd.get_dummies(df, columns=['color'])

# Handle dates
df['date'] = pd.to_datetime(df['date'])
df['day_of_week'] = df['date'].dt.day_name()