Boolesche Indexierung auf DataFrames

Boolean indexing lets you filter rows in a DataFrame by applying a condition that produces a Series of True and False values. Only rows where the condition is True are returned. This is one of the most frequently used selection techniques in Pandas because it is both readable and fast.

For example, given a sales DataFrame, you can instantly retrieve all rows where revenue exceeded 1000 without writing a loop.

Once you have a boolean mask, you pass it inside square brackets on the DataFrame to select only the matching rows. The result is a new DataFrame — the original is never modified. The row indices from the original are preserved, so you always know where each row came from.

This pattern — create mask, then apply — is the standard Pandas idiom for row filtering.

You do not need to assign the boolean Series to a variable. You can write the condition inline directly inside the brackets: df[df['col'] > value]. This one-liner style is very common in data analysis notebooks because it keeps code compact and readable.

Both approaches — mask variable and inline — produce identical results. Use a variable when the condition is complex or reused; use inline for simple one-off filters.

To require that both conditions are true, combine them with the & operator — not Python's and keyword, which does not work element-wise on arrays. Each condition must be wrapped in parentheses because & has higher operator precedence than comparison operators.

The result keeps only rows where every sub-condition evaluates to True.

Use the | operator to keep rows where at least one condition is true. As with &, each sub-condition must be in parentheses. The | operator performs element-wise logical OR across the boolean arrays.

Mixing & and | is perfectly valid — just be careful with parentheses to enforce the correct evaluation order and avoid subtle logic bugs.

The tilde ~ operator inverts a boolean Series — it turns True into False and vice versa. Use ~ to express 'NOT' logic: keep rows that do not match a condition. This is cleaner than constructing the opposite condition manually.

For example, df[~df['status'].isin(['cancelled', 'refunded'])] keeps all rows except those two statuses.

Boolean indexing works just as well on text columns. You can filter for an exact match with ==, or use .str methods like .str.startswith(), .str.contains(), or .str.upper() inside the condition for flexible text filtering.

String comparisons in Pandas are case-sensitive by default, so 'NYC' and 'nyc' are treated as different values. Normalise case first if needed.

Complex analyses often require conditions across multiple columns combined with & and |. Break very long conditions into named boolean Series for readability — each acting as a named sub-filter that you combine at the end.

This approach makes your intent clear: each line reads like an English sentence describing one part of the filter logic.

Boolean indexing filters rows, but sometimes you want to add a new column based on a condition rather than drop rows. np.where(condition, value_if_true, value_if_false) is the vectorised equivalent of an if/else applied element-wise across a column, and it is much faster than using apply with a lambda.

You can update values in place for filtered rows using .loc[mask, column]. This is the safe way to set values on a subset — using chained indexing like df[mask]['col'] = value may raise a SettingWithCopyWarning because it operates on a copy.

Always prefer df.loc[mask, 'col'] = value when you want to modify the original DataFrame in place.

Since boolean values are 1 (True) and 0 (False) under the hood, you can call .sum() on a mask to count matching rows, or .mean() to get the fraction of rows that match. These quick counts help you verify your filter logic before applying it to select rows.

Test your understanding of boolean indexing on DataFrames.

In this lesson you learned: boolean masks filter rows by applying a condition to a column, & and | operators combine multiple conditions (not Python's and/or), and ~ inverts a boolean mask for NOT logic. Use df.loc[mask, col] = value to safely assign values to filtered rows. Next up we explore the query() method for writing filters as readable strings.