fs.load()

When to Use It

Use at the start of any data validation script or pipeline step to parse files or wrap DataFrames into a standard Dataset object containing inferred schemas, column data types, row counts, and source metadata.

Arguments

• source: str | pandas.DataFrame | polars.DataFrame. Local file path (.csv, .xlsx, .xls, .parquet) or loaded DataFrame object.

Return Value

Returns a normalized, shallowly immutable Dataset dataclass containing dataframe, backend ("polars" or "pandas"), schema (DatasetSchema), row_count, column_count, dtypes, source, and file_size.

Exceptions

• ConnectorError: Base exception raised when a data source cannot be validated or loaded.
• SourceNotFoundError: Raised when the target local file path does not exist.
• UnsupportedFormatError: Raised when the file extension or object type is unsupported.
• SourceParseError: Raised when parsing or reading the file content fails.

Example

1import featuresmith as fs
2import polars as pl
3
4# Load from local file path (CSV, Parquet, Excel)
5ds = fs.load("train.parquet")
6print(f"Loaded {ds.row_count} rows across {ds.column_count} columns via {ds.backend}.")
7
8# Load from in-memory Polars or pandas DataFrame
9df = pl.DataFrame({"x": [1, 2, 3], "y": [4.0, 5.0, 6.0]})
10ds_mem = fs.load(df)
11print(ds_mem.preview(2))

Notes and Limitations

• Zero Data Copying: In-memory pandas or Polars DataFrames are wrapped directly without copying memory buffers.
• Backend Engines: Polars is used for CSV and Parquet files; pandas is used for Excel files.