Threshold-based cleaning
Sanitize tabular measurements by replacing values that violate a threshold.
Pandas · Module
df.mask
DataFrame.mask replaces values where a condition is True and keeps values where it is False. Replacement values come from other, or from dtype-appropriate nulls when other is omitted.
How it works
DataFrame.mask evaluates a boolean condition element by element. Values where the condition is False remain unchanged; values where it is True come from other, or become the dtype-appropriate missing value when other is omitted. The operation returns a DataFrame unless inplace=True.
Syntax
DataFrame.mask(cond, other=<no_default>, *, inplace=False, axis=None, level=None)Parameters
| Name | Type | Description |
|---|---|---|
cond | bool Series/DataFrame, array-like, or callable | Boolean Series/DataFrame, array-like, or callable selecting entries to replace; False entries are retained. |
other | scalar, Series/DataFrame, or callable | Scalar, Series/DataFrame, or callable supplying replacement values; omitted values become dtype-appropriate nulls. |
inplace | bool, default False | Whether to perform the operation in place. |
axis | int, default None | Alignment axis when needed. |
level | int, default None | Alignment level when needed. |
Returns
A DataFrame when called on a DataFrame; None when inplace=True.
Key points
- mask selects positions whose condition evaluates to True and replaces only those positions.
- False condition entries retain their original DataFrame values.
- A callable condition is evaluated on the DataFrame and must return boolean-shaped output.
- The replacement can come from a scalar or corresponding values in another aligned object.
- Leaving other unspecified fills selected entries with a dtype-appropriate null.
- Set inplace=True to request mutation instead of receiving a returned DataFrame.
Common mistakes
- A callable supplied as cond or other must not mutate the DataFrame it receives.
- Misaligned condition and replacement labels can cause those positions to be replaced, so align inputs deliberately.
- The original object's dtype influences how replacement values are represented.
Example: Replace values matching a condition
import pandas as pd
scores = pd.DataFrame({"math": [72, 91], "science": [88, 64]})
result = scores.mask(scores < 70, other=0)
print(result)Replaces every score below 70 with zero while leaving all other values unchanged.
When to use it
Conditional redaction
Redact selected cells while preserving the rest of a DataFrame.
Conditional replacement
Apply a scalar, aligned object, or computed replacement to matching entries.
Interview questions
QWhat does DataFrame.mask do?
mask replaces elements where cond is True and preserves elements where cond is False. With a DataFrame input, it returns a DataFrame unless inplace=True.
Follow-up: How does mask differ from where? where replaces values where the condition is False, while mask replaces values where it is True.
QWhat forms can cond take?
cond can be a boolean Series, DataFrame, array-like object, or callable. A callable is evaluated against the DataFrame and must produce a boolean Series, DataFrame, or array.
Follow-up: What restriction applies to that callable? It must not modify the input DataFrame, even though pandas does not validate this.
QHow is the replacement value chosen?
other may be a scalar, Series, DataFrame, or callable. If omitted, positions selected by cond are filled with the appropriate null value, such as np.nan for NumPy dtypes or pd.NA for extension dtypes.
Follow-up: What happens when replacement labels do not align with the condition? Misaligned condition positions are treated as True and therefore are replaced.
Related
DataFrame.whereSeries.maskSeries.where