Column cardinality checks
Profile categorical or identifier columns by measuring how many different values each contains.
Pandas · Module
df.nunique
DataFrame.nunique counts distinct values along an axis and returns a Series. It excludes NaN by default, and can count per row with axis=1 or include missing values with dropna=False.
How it works
DataFrame.nunique counts distinct values along a chosen axis and returns the counts as a Series. By default it excludes missing values; set dropna=False to include them, and use axis=1 or "columns" to count across each row.
Syntax
DataFrame.nunique(axis=0, dropna=True)Parameters
| Name | Type | Description |
|---|---|---|
axis | {0 or 'index', 1 or 'columns'}, default 0 | Axis to use: 0 or "index" counts down each column; 1 or "columns" counts across each row. Defaults to 0. |
dropna | bool, default True | Whether to include NaN values in the distinct counts. Defaults to True, meaning they are excluded. |
Returns
Series with counts of unique values per row or column, depending on axis.
Key points
- The method measures distinct elements along the selected DataFrame axis.
- The output is a Series of counts rather than another DataFrame.
- The default axis is 0, which yields one count for each column.
- dropna defaults to True, so missing values are not counted unless requested.
- The returned Series represents per-column or per-row counts according to axis.
Common mistakes
- Missing values are excluded by default; pass dropna=False when they should contribute to the distinct-value count.
- The axis labels describe the counted direction, so axis=0 or "index" produces per-column counts, while axis=1 or "columns" produces per-row counts.
Example: Count distinct values by column or row
import pandas as pd
scores = pd.DataFrame({
"math": [90, 90, 80],
"english": [85, 85, None],
})
by_column = scores.nunique()
by_row_including_missing = scores.nunique(axis="columns", dropna=False)
print(by_column)
print(by_row_including_missing)Counts distinct values for each column and then for each row, treating the missing value as a distinct value in the second calculation.
When to use it
Row-level uniqueness analysis
Detect rows containing varied values by counting distinct entries across each row.
Missing-value-aware profiling
Include missingness as part of a row or column's distinct-value profile when that distinction matters.
Interview questions
QWhat does DataFrame.nunique return by default?
DataFrame.nunique returns a Series containing the number of distinct values for each column by default. Its default axis is 0, and missing values are omitted unless dropna is set to False.
Follow-up: How would you count distinct values across rows? Use axis=1 or the equivalent axis="columns".
QHow do you count unique values row by row?
Use axis=1, or axis="columns", to count unique values within each row. The result is a Series indexed by the DataFrame's row labels.
Follow-up: How can missing values affect that result? They are ignored by default; pass dropna=False to include them.
QWhich options control DataFrame.nunique behavior?
The method accepts axis values 0 or "index" and 1 or "columns". The dropna boolean controls whether NaN values participate in the count and defaults to True.
Follow-up: What type is returned for either direction? A Series containing counts per column or per row, depending on the selected axis.
Related
Series.nuniqueDataFrame.count