fs.profile()

When to Use It

Use when you need detailed statistical profiles of dataset columns (min, max, mean, quantiles, missingness, cardinality, frequency tables, correlations) without running rule evaluations or review scoring.

Arguments

• source: Dataset | str | DataFrame. Pre-loaded Dataset or file/data source.
• max_correlation_columns: int (default 100). Column cap for Pearson correlation computations to prevent combinatorial blowup.
• max_frequency_table_size: int (default 1000). Maximum unique categories to track in frequency table summaries.

Return Value

Returns a frozen ProfileResult dataclass containing dataset_summary, column_profiles, typed profiles (numeric_profiles, categorical_profiles, datetime_profiles, text_profiles), missingness & duplicate summaries, correlation matrices, and execution metadata.

Exceptions

• ConnectorError: Raised if an unresolved file path or invalid DataFrame source fails to load before profiling.

Example

1import featuresmith as fs
2
3profile = fs.profile("customers.csv", max_correlation_columns=50)
4
5# Inspect column summaries
6print(profile.column_profiles["age"].missing_count)
7print(profile.dataset_summary.row_count)
8print(profile.numeric_profiles["income"].mean)

Notes and Limitations

• Deterministic Engine: Computations run on Polars or pandas backend using vectorized primitives.
• Frozen Output: The returned ProfileResult is fully frozen, slotted, and serializable via profile.to_dict().