featuresmith-core 0.1.0__py3-none-any.whl

featuresmith/__init__.py

@@ -0,0 +1,7 @@
+"""Featuresmith core package."""
+
+from featuresmith.api import analyze, load, profile
+
+__version__ = "0.1.0"
+
+__all__ = ["analyze", "load", "profile"]

featuresmith/api.py

@@ -0,0 +1,180 @@
+"""Public SDK entry module for Featuresmith."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from featuresmith.connectors.registry import default_registry
+from featuresmith.core.dataset import Dataset as Dataset
+from featuresmith.core.exceptions import (
+    ConnectorError as ConnectorError,
+)
+from featuresmith.core.exceptions import (
+    SourceNotFoundError as SourceNotFoundError,
+)
+from featuresmith.core.exceptions import (
+    SourceParseError as SourceParseError,
+)
+from featuresmith.core.exceptions import (
+    UnsupportedFormatError as UnsupportedFormatError,
+)
+from featuresmith.core.profile_result import ProfileResult as ProfileResult
+from featuresmith.core.rule_result import RuleResult as RuleResult
+from featuresmith.profiling import profile_dataset
+
+
+def load(source: object) -> Dataset:
+    """Load a supported tabular source into a normalized Dataset.
+
+    Args:
+        source: The data source to load. This can be a string representing a
+            local file path (CSV, Excel, Parquet), or an in-memory pandas
+            DataFrame or Polars DataFrame.
+
+    Returns:
+        Dataset: A normalized view of the loaded tabular dataset, containing
+            the dataframe backend, schema, dtypes, and metadata.
+
+    Raises:
+        ConnectorError: If the source is missing, has an unsupported format,
+            is corrupted, or is of an invalid type.
+
+    Notes:
+        For file paths, Featuresmith utilizes Polars for CSV and Parquet formats
+        by default, and pandas for Excel formats. In-memory dataframes are
+        wrapped without copying the underlying memory.
+
+    Examples:
+        >>> import pandas as pd
+        >>> import featuresmith as fs
+        >>> df = pd.DataFrame({"a": [1, 2, 3]})
+        >>> dataset = fs.load(df)
+        >>> dataset.row_count
+        3
+    """
+    return default_registry().load(source)
+
+
+def profile(
+    source: object,
+    *,
+    max_correlation_columns: int = 100,
+    max_frequency_table_size: int = 1000,
+) -> ProfileResult:
+    """Profile a supported source or Dataset and compute statistical summaries.
+
+    Args:
+        source: The data source to profile. This can be a pre-loaded Dataset
+            object, a string representing a local file path (CSV, Excel,
+            Parquet), or an in-memory pandas or Polars DataFrame.
+        max_correlation_columns: Limit correlation computations to prevent
+            combinatorial blowup (default 100).
+        max_frequency_table_size: Maximum entries to keep in categorical
+            frequency tables (default 1000).
+
+    Returns:
+        ProfileResult: A strongly-typed statistical profile containing dataset
+            summaries, column profiles, correlations, and logical types.
+
+    Raises:
+        ConnectorError: If the source is a file path or dataframe that fails to
+            load before profiling.
+
+    Notes:
+        This is a deterministic profiling engine. The statistical calculations
+        run on the underlying dataframe backend using vectorized Polars or pandas
+        APIs. The resulting ProfileResult is frozen and fully serializable.
+
+    Examples:
+        >>> import polars as pl
+        >>> import featuresmith as fs
+        >>> df = pl.DataFrame({"a": [1.0, 2.0, None]})
+        >>> prof = fs.profile(df)
+        >>> prof.dataset_summary.column_count
+        1
+        >>> prof.column_profiles["a"].missing_count
+        1
+    """
+    if isinstance(source, Dataset):
+        dataset = source
+    else:
+        dataset = load(source)
+    return profile_dataset(
+        dataset,
+        max_correlation_columns=max_correlation_columns,
+        max_frequency_table_size=max_frequency_table_size,
+    )
+
+
+def analyze(
+    source: object,
+    *,
+    target_column: str | None = None,
+    enabled_rules: list[str] | None = None,
+    rule_config: dict[str, Any] | None = None,
+    max_correlation_columns: int = 100,
+    max_frequency_table_size: int = 1000,
+) -> RuleResult:
+    """Analyze a tabular source or Dataset, computing profile stats and running rules.
+
+    Args:
+        source: The data source to analyze. This can be a pre-loaded Dataset
+            object, a string representing a local file path (CSV, Excel,
+            Parquet), or an in-memory pandas or Polars DataFrame.
+        target_column: Optional name of the target column in the dataset, used
+            specifically for evaluating potential target leakage.
+        enabled_rules: Optional list of rule IDs to execute. If not provided,
+            the engine runs all rules that are enabled by default.
+        rule_config: Optional dictionary of configurations keyed by rule ID.
+            For example, `{"quality.missing_value_threshold": {"threshold": 15.0}}`.
+        max_correlation_columns: Limit correlation computations during profiling to
+            prevent combinatorial blowup (default 100).
+        max_frequency_table_size: Maximum entries to keep in categorical
+            frequency tables (default 1000).
+
+    Returns:
+        RuleResult: The canonical output of the Rule Engine containing the
+            computed ProfileResult, aggregated list of RuleFindings, list of
+            executed rule IDs, and execution time metadata.
+
+    Raises:
+        ConnectorError: If the source is a file path or dataframe that fails to
+            load before profiling.
+
+    Notes:
+        This function integrates connector loading, profiling, and rule
+        evaluation into a single public endpoint. The evaluation of rules is
+        deterministic and isolated; a crash in any single rule does not cause
+        the function to fail, but is recorded in the failed_rules mapping.
+
+    Examples:
+        >>> import pandas as pd
+        >>> import featuresmith as fs
+        >>> df = pd.DataFrame({
+        ...     "x": [1, 2, 3, 4, 5],
+        ...     "y": [1.0, 2.0, 3.0, 4.0, 5.0],
+        ...     "target": [0, 1, 0, 1, 0]
+        ... })
+        >>> result = fs.analyze(df, target_column="target")
+        >>> len(result.findings) >= 0
+        True
+    """
+    if isinstance(source, Dataset):
+        dataset = source
+    else:
+        dataset = load(source)
+    prof_res = profile(
+        dataset,
+        max_correlation_columns=max_correlation_columns,
+        max_frequency_table_size=max_frequency_table_size,
+    )
+
+    from featuresmith.rules.engine import RuleEngine
+
+    engine = RuleEngine()
+    return engine.run(
+        prof_res,
+        target_column=target_column,
+        enabled_rules=enabled_rules,
+        rule_config=rule_config,
+    )

featuresmith/connectors/__init__.py

@@ -0,0 +1,18 @@
+"""Connector package for Featuresmith."""
+
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.connectors.csv_connector import CsvConnector
+from featuresmith.connectors.dataframe_connector import DataFrameConnector
+from featuresmith.connectors.excel_connector import ExcelConnector
+from featuresmith.connectors.parquet_connector import ParquetConnector
+from featuresmith.connectors.registry import ConnectorRegistry, default_registry
+
+__all__ = [
+    "BaseConnector",
+    "ConnectorRegistry",
+    "CsvConnector",
+    "DataFrameConnector",
+    "ExcelConnector",
+    "ParquetConnector",
+    "default_registry",
+]

featuresmith/connectors/_paths.py

@@ -0,0 +1,25 @@
+"""Private helpers shared by local-file connectors."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from featuresmith.core.exceptions import SourceNotFoundError, UnsupportedFormatError
+
+
+def validate_file_source(source: object, suffixes: tuple[str, ...]) -> Path:
+    """Return a validated local file path for one of the supplied suffixes."""
+    if not isinstance(source, (str, Path)):
+        raise UnsupportedFormatError("Expected a local file path.")
+
+    path = Path(source)
+    if path.suffix.lower() not in suffixes:
+        formats = ", ".join(suffixes)
+        raise UnsupportedFormatError(
+            f"Unsupported file format. Expected one of: {formats}."
+        )
+    if not path.exists():
+        raise SourceNotFoundError(f"Source file does not exist: {path}.")
+    if not path.is_file():
+        raise SourceNotFoundError(f"Source path is not a file: {path}.")
+    return path

featuresmith/connectors/base.py

@@ -0,0 +1,54 @@
+"""Base contract for source connectors."""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+
+from featuresmith.core.dataset import Dataset
+
+
+class BaseConnector(ABC):
+    """Abstract base class for all Featuresmith data connectors.
+
+    Connectors are responsible for loading and normalizing external tabular
+    sources (such as files or database connections) or in-memory representations
+    into a standardized Dataset object.
+    """
+
+    @abstractmethod
+    def can_load(self, source: object) -> bool:
+        """Determine whether the connector supports the supplied source.
+
+        Args:
+            source: The input source object (e.g. file path or DataFrame).
+
+        Returns:
+            bool: True if this connector supports loading from the source,
+                False otherwise.
+        """
+
+    @abstractmethod
+    def validate(self, source: object) -> None:
+        """Validate the source structurally or raise a typed connector error.
+
+        Args:
+            source: The input source object to validate.
+
+        Raises:
+            ConnectorError: If validation fails because the source is missing,
+                unsupported, or invalid.
+        """
+
+    @abstractmethod
+    def load(self, source: object) -> Dataset:
+        """Load and normalize a validated source into a Dataset.
+
+        Args:
+            source: The input source object to load.
+
+        Returns:
+            Dataset: The loaded and normalized dataset wrapper.
+
+        Raises:
+            ConnectorError: If loading fails due to parsing or read errors.
+        """

featuresmith/connectors/csv_connector.py

@@ -0,0 +1,44 @@
+"""CSV connector backed by Polars."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+import polars as pl
+
+from featuresmith.connectors._paths import validate_file_source
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import SourceParseError
+
+
+class CsvConnector(BaseConnector):
+    """Load local CSV files into Polars-backed datasets."""
+
+    _SUFFIXES = (".csv",)
+
+    def can_load(self, source: object) -> bool:
+        """Return whether the source is a CSV path."""
+        return (
+            isinstance(source, (str, Path))
+            and Path(source).suffix.lower() in self._SUFFIXES
+        )
+
+    def validate(self, source: object) -> None:
+        """Validate that the CSV source exists and is a regular file."""
+        validate_file_source(source, self._SUFFIXES)
+
+    def load(self, source: object) -> Dataset:
+        """Load a CSV file as a normalized Polars dataset."""
+        path = validate_file_source(source, self._SUFFIXES)
+        try:
+            dataframe = pl.read_csv(path)
+        except (OSError, pl.exceptions.PolarsError) as error:
+            raise SourceParseError(f"Could not read CSV file '{path}'.") from error
+        return Dataset.from_dataframe(
+            dataframe,
+            backend="polars",
+            source=str(path),
+            file_size=path.stat().st_size,
+            metadata={"format": "csv"},
+        )

featuresmith/connectors/dataframe_connector.py

@@ -0,0 +1,40 @@
+"""In-memory pandas and Polars dataframe connector."""
+
+from __future__ import annotations
+
+import pandas as pd
+import polars as pl
+
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import UnsupportedFormatError
+
+
+class DataFrameConnector(BaseConnector):
+    """Wrap in-memory pandas or Polars dataframes without copying them."""
+
+    def can_load(self, source: object) -> bool:
+        """Return whether the source is a supported dataframe instance."""
+        return isinstance(source, (pd.DataFrame, pl.DataFrame))
+
+    def validate(self, source: object) -> None:
+        """Validate that the source is a pandas or Polars dataframe."""
+        if not self.can_load(source):
+            raise UnsupportedFormatError("Expected a pandas or Polars DataFrame.")
+
+    def load(self, source: object) -> Dataset:
+        """Wrap an in-memory dataframe in the normalized dataset contract."""
+        self.validate(source)
+        if isinstance(source, pd.DataFrame):
+            return Dataset.from_dataframe(
+                source,
+                backend="pandas",
+                metadata={"format": "dataframe"},
+            )
+        if isinstance(source, pl.DataFrame):
+            return Dataset.from_dataframe(
+                source,
+                backend="polars",
+                metadata={"format": "dataframe"},
+            )
+        raise UnsupportedFormatError("Expected a pandas or Polars DataFrame.")

featuresmith/connectors/excel_connector.py

@@ -0,0 +1,45 @@
+"""Excel connector backed by pandas."""
+
+from __future__ import annotations
+
+import zipfile
+from pathlib import Path
+
+import pandas as pd
+
+from featuresmith.connectors._paths import validate_file_source
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import SourceParseError
+
+
+class ExcelConnector(BaseConnector):
+    """Load the first worksheet from a local Excel workbook."""
+
+    _SUFFIXES = (".xlsx", ".xls", ".xlsm")
+
+    def can_load(self, source: object) -> bool:
+        """Return whether the source is an Excel workbook path."""
+        return (
+            isinstance(source, (str, Path))
+            and Path(source).suffix.lower() in self._SUFFIXES
+        )
+
+    def validate(self, source: object) -> None:
+        """Validate that the Excel source exists and is a regular file."""
+        validate_file_source(source, self._SUFFIXES)
+
+    def load(self, source: object) -> Dataset:
+        """Load the first worksheet as a normalized pandas dataset."""
+        path = validate_file_source(source, self._SUFFIXES)
+        try:
+            dataframe = pd.read_excel(path)
+        except (ImportError, OSError, ValueError, zipfile.BadZipFile) as error:
+            raise SourceParseError(f"Could not read Excel file '{path}'.") from error
+        return Dataset.from_dataframe(
+            dataframe,
+            backend="pandas",
+            source=str(path),
+            file_size=path.stat().st_size,
+            metadata={"format": "excel", "sheet": 0},
+        )

featuresmith/connectors/parquet_connector.py

@@ -0,0 +1,44 @@
+"""Parquet connector backed by Polars."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+import polars as pl
+
+from featuresmith.connectors._paths import validate_file_source
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import SourceParseError
+
+
+class ParquetConnector(BaseConnector):
+    """Load local Parquet files into Polars-backed datasets."""
+
+    _SUFFIXES = (".parquet", ".pq")
+
+    def can_load(self, source: object) -> bool:
+        """Return whether the source is a Parquet path."""
+        return (
+            isinstance(source, (str, Path))
+            and Path(source).suffix.lower() in self._SUFFIXES
+        )
+
+    def validate(self, source: object) -> None:
+        """Validate that the Parquet source exists and is a regular file."""
+        validate_file_source(source, self._SUFFIXES)
+
+    def load(self, source: object) -> Dataset:
+        """Load a Parquet file as a normalized Polars dataset."""
+        path = validate_file_source(source, self._SUFFIXES)
+        try:
+            dataframe = pl.read_parquet(path)
+        except (OSError, pl.exceptions.PolarsError) as error:
+            raise SourceParseError(f"Could not read Parquet file '{path}'.") from error
+        return Dataset.from_dataframe(
+            dataframe,
+            backend="polars",
+            source=str(path),
+            file_size=path.stat().st_size,
+            metadata={"format": "parquet"},
+        )

featuresmith/connectors/registry.py

@@ -0,0 +1,55 @@
+"""Static connector registration and source dispatch for Sprint 2."""
+
+from __future__ import annotations
+
+from collections.abc import Iterable
+
+from featuresmith.connectors.base import BaseConnector
+from featuresmith.connectors.csv_connector import CsvConnector
+from featuresmith.connectors.dataframe_connector import DataFrameConnector
+from featuresmith.connectors.excel_connector import ExcelConnector
+from featuresmith.connectors.parquet_connector import ParquetConnector
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import UnsupportedFormatError
+
+
+class ConnectorRegistry:
+    """Select a registered connector for a supported input source.
+
+    Registration is intentionally explicit during Sprint 2. It establishes the
+    extension boundary without entry-point discovery or dynamic loading.
+    """
+
+    def __init__(self, connectors: Iterable[BaseConnector] = ()) -> None:
+        """Create a registry with optional initial connector instances."""
+        self._connectors = list(connectors)
+
+    def register(self, connector: BaseConnector) -> None:
+        """Register a connector instance for subsequent source dispatch."""
+        self._connectors.append(connector)
+
+    def load(self, source: object) -> Dataset:
+        """Load a source with the first registered connector that supports it.
+
+        Raises:
+            ConnectorError: If no registered connector supports the source.
+        """
+        for connector in self._connectors:
+            if connector.can_load(source):
+                return connector.load(source)
+        raise UnsupportedFormatError(
+            "Unsupported source. Use a CSV, Excel, Parquet, pandas DataFrame, "
+            "or Polars DataFrame."
+        )
+
+
+def default_registry() -> ConnectorRegistry:
+    """Return the built-in Sprint 2 connector registry."""
+    return ConnectorRegistry(
+        (
+            DataFrameConnector(),
+            CsvConnector(),
+            ExcelConnector(),
+            ParquetConnector(),
+        )
+    )

featuresmith/core/__init__.py

@@ -0,0 +1,50 @@
+"""Core primitives for Featuresmith."""
+
+from featuresmith.core.dataset import Dataset
+from featuresmith.core.exceptions import (
+    ConnectorError,
+    SourceNotFoundError,
+    SourceParseError,
+    UnsupportedFormatError,
+)
+from featuresmith.core.profile_result import (
+    CategoricalProfile,
+    ColumnProfile,
+    CorrelationSummary,
+    DatasetMetadata,
+    DatasetSummary,
+    DatetimeProfile,
+    DuplicateSummary,
+    ExecutionMetadata,
+    MissingValueSummary,
+    NumericProfile,
+    ProfileResult,
+    TextProfile,
+)
+from featuresmith.core.rule_finding import RuleFinding
+from featuresmith.core.rule_result import RuleResult
+from featuresmith.core.schema import ColumnSchema, DatasetSchema
+
+__all__ = [
+    "ColumnSchema",
+    "ConnectorError",
+    "SourceNotFoundError",
+    "SourceParseError",
+    "UnsupportedFormatError",
+    "Dataset",
+    "DatasetSchema",
+    "ColumnProfile",
+    "CategoricalProfile",
+    "CorrelationSummary",
+    "DatasetMetadata",
+    "DatasetSummary",
+    "DatetimeProfile",
+    "DuplicateSummary",
+    "ExecutionMetadata",
+    "MissingValueSummary",
+    "NumericProfile",
+    "ProfileResult",
+    "TextProfile",
+    "RuleFinding",
+    "RuleResult",
+]

featuresmith/core/dataset.py

@@ -0,0 +1,107 @@
+"""Normalized dataset abstraction passed between Featuresmith pipeline stages."""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from dataclasses import dataclass, field
+from types import MappingProxyType
+from typing import Any
+
+from featuresmith.core.schema import ColumnSchema, DatasetSchema
+
+
+@dataclass(frozen=True, slots=True)
+class Dataset:
+    """A lightweight, normalized view of a loaded tabular dataset.
+
+    The wrapper is shallowly immutable: its descriptive fields cannot be
+    reassigned, while the underlying dataframe remains owned by its backend.
+    Connectors construct this object; it deliberately contains no profiling or
+    source-specific behavior.
+
+    Attributes:
+        dataframe: The loaded pandas or Polars dataframe.
+        backend: The dataframe backend identifier, either ``"pandas"`` or
+            ``"polars"``.
+        schema: Ordered schema inferred from the dataframe.
+        metadata: Read-only descriptive metadata supplied by the connector.
+        row_count: Number of rows in the dataframe.
+        column_count: Number of columns in the dataframe.
+        dtypes: Read-only mapping from column name to dtype string.
+        source: Original file path when loaded from a file, otherwise ``None``.
+        file_size: File size in bytes when loaded from a file, otherwise
+            ``None``.
+    """
+
+    dataframe: Any
+    backend: str
+    schema: DatasetSchema
+    metadata: Mapping[str, object] = field(default_factory=dict)
+    row_count: int = 0
+    column_count: int = 0
+    dtypes: Mapping[str, str] = field(default_factory=dict)
+    source: str | None = None
+    file_size: int | None = None
+
+    def __post_init__(self) -> None:
+        """Freeze mapping fields so dataset descriptors stay immutable."""
+        object.__setattr__(self, "metadata", MappingProxyType(dict(self.metadata)))
+        object.__setattr__(self, "dtypes", MappingProxyType(dict(self.dtypes)))
+
+    @classmethod
+    def from_dataframe(
+        cls,
+        dataframe: Any,
+        *,
+        backend: str,
+        source: str | None = None,
+        file_size: int | None = None,
+        metadata: Mapping[str, object] | None = None,
+    ) -> Dataset:
+        """Create a normalized dataset from a dataframe backend object.
+
+        Args:
+            dataframe: A pandas or Polars dataframe.
+            backend: Identifier for the dataframe backend.
+            source: Original local file path, if any.
+            file_size: Source file size in bytes, if known.
+            metadata: Connector-provided descriptive metadata.
+
+        Returns:
+            A normalized dataset with an inferred schema and dtype mapping.
+        """
+        columns = tuple(str(column) for column in dataframe.columns)
+        dtype_values = tuple(str(dtype) for dtype in dataframe.dtypes)
+        dtypes = dict(zip(columns, dtype_values, strict=True))
+        schema = DatasetSchema(
+            columns=tuple(
+                ColumnSchema(name=name, dtype=dtype) for name, dtype in dtypes.items()
+            )
+        )
+        return cls(
+            dataframe=dataframe,
+            backend=backend,
+            schema=schema,
+            metadata=metadata or {},
+            row_count=len(dataframe),
+            column_count=len(columns),
+            dtypes=dtypes,
+            source=source,
+            file_size=file_size,
+        )
+
+    def preview(self, rows: int = 5) -> Any:
+        """Return the first requested number of rows from the dataframe.
+
+        Args:
+            rows: Number of rows to return. Must be non-negative.
+
+        Returns:
+            A dataframe of the same backend containing the requested head rows.
+
+        Raises:
+            ValueError: If ``rows`` is negative.
+        """
+        if rows < 0:
+            raise ValueError("Preview row count must be non-negative.")
+        return self.dataframe.head(rows)