Indexing and Selection¶

Indexing is one of pandas' most powerful features. Understanding the difference between label-based and position-based selection is essential.

Label-based Selection¶

The loc accessor selects by labels.

1. Single Row by Label¶

import pandas as pd

df = pd.DataFrame({
    "price": [100, 101, 102],
    "volume": [10, 12, 9]
}, index=['day1', 'day2', 'day3'])

print(df.loc['day1'])

price     100
volume     10
Name: day1, dtype: int64

2. Row and Column Selection¶

df.loc['day1', 'price']       # Single value
df.loc['day1':'day2', 'price']  # Slice (inclusive)
df.loc[:, 'price']            # All rows, one column

3. Multiple Labels¶

df.loc[['day1', 'day3'], ['price', 'volume']]

Position-based Selection¶

The iloc accessor selects by integer positions.

1. Single Row by Position¶

df.iloc[0]  # First row

2. Row and Column by Position¶

df.iloc[0, 0]      # First row, first column
df.iloc[0:2, 0]    # First two rows, first column
df.iloc[:, 0]      # All rows, first column

3. Integer Slicing¶

df.iloc[0:2]       # First two rows (exclusive end)
df.iloc[-1]        # Last row

Boolean Indexing¶

Select rows based on conditions.

1. Single Condition¶

df[df["price"] > 100]

2. Multiple Conditions¶

df[(df["price"] > 100) & (df["volume"] > 10)]
df[(df["price"] > 102) | (df["volume"] < 10)]

3. Using Query¶

df.query("price > 100 and volume > 10")

Chained Indexing¶

Avoid chained indexing to prevent unexpected behavior.

1. Problematic Pattern¶

# This can cause SettingWithCopyWarning
df[df["price"] > 100]["volume"] = 20

2. Correct Approach¶

df.loc[df["price"] > 100, "volume"] = 20

3. Copy vs View¶

Chained indexing may return a view or copy unpredictably.

Setting Values¶

Use loc and iloc for assignment.

1. Single Value¶

df.loc['day1', 'price'] = 105

2. Multiple Values¶

df.loc['day1', ['price', 'volume']] = [105, 15]

3. Conditional Assignment¶

df.loc[df['price'] > 100, 'flag'] = True

Best Practices¶

Follow these guidelines for clean indexing code.

1. Explicit Selection¶

Always use loc or iloc explicitly:

# Preferred
df.loc[0]     # If 0 is a label
df.iloc[0]    # If 0 is a position

# Avoid
df[0]         # Ambiguous

2. Avoid Mixed Indexing¶

# Don't mix labels and positions
# Use loc for labels, iloc for positions

3. Check Index Type¶

print(df.index)       # Understand your index
print(type(df.index)) # Know the index type