Beyond single labels and slices, .loc is very flexible with how you select columns. Here are the other "selectors" you can use in the column_selector position:
1. A List of Labels
Unlike a slice ('name':'major') which selects everything in between, a list allows you to pick specific, non-adjacent columns.
# Select names and scores, skipping age and major
df.loc[1:3, ['name', 'score']]2. A Colon : (Select All)
If you want to filter specific rows based on a condition but keep all columns, use a single colon.
# Select rows where age > 21, and keep every column
df.loc[df['age'] > 21, :]3. Boolean Arrays
You can pass a list of True and False values that matches the number of columns. This is often used with logic.
# Select only columns that start with the letter 'm'
column_mask = [col.startswith('m') for col in df.columns]
df.loc[:, column_mask]4. A Callable (Function)
You can pass a lambda function that looks at the DataFrame. This is useful for "chaining" operations.
# Select columns where the column name has more than 4 characters
df.loc[:, lambda x: [col for col in x.columns if len(col) > 4]]Summary Table for .loc[row_selector, column_selector]
| Selector Type | Example | Result |
|---|---|---|
| Single Label | 'name' |
Returns a Series (one column) |
| List of Labels | ['name', 'score'] |
Returns specific columns |
| Slice | 'name':'major' |
Returns columns in that range (inclusive) |
| Colon | : |
Returns all columns |
| Boolean Array | [True, False, True, False] |
Returns columns where value is True |
Which of these would be most useful for the data you are currently working with?
