snowflake-ml-python 1.22.0__py3-none-any.whl → 1.24.0__py3-none-any.whl

snowflake/ml/feature_store/feature.py

@@ -0,0 +1,366 @@
+"""User-facing Feature builder for defining feature view features.
+
+This module provides the Feature class for defining features in a FeatureView,
+including both aggregated features (for tiled feature views) and non-aggregated
+fields (for standard feature views).
+"""
+
+from __future__ import annotations
+
+from typing import Any, Optional
+
+from snowflake.ml.feature_store.aggregation import (
+    AggregationSpec,
+    AggregationType,
+    is_lifetime_window,
+)
+
+
+class Feature:
+    """Fluent builder for defining features in a FeatureView.
+
+    This class provides a user-friendly API for defining features with
+    time-series aggregations. It supports method chaining for setting
+    options like aliases.
+
+    Example::
+
+        >>> from snowflake.ml.feature_store import Feature
+        >>>
+        >>> # Define features with factory methods
+        >>> amount_sum = Feature.sum("amount", "24h").alias("total_amount_24h")
+        >>> recent_pages = Feature.last_n("page_id", "1h", n=10).alias("recent_pages")
+        >>> txn_count = Feature.count("transaction_id", "7d")
+    """
+
+    def __init__(
+        self,
+        function: AggregationType,
+        column: str,
+        window: str,
+        offset: str = "0",
+        **params: Any,
+    ) -> None:
+        """Create a Feature with the specified aggregation.
+
+        Args:
+            function: The aggregation function type.
+            column: The source column to aggregate.
+            window: The lookback window (e.g., "24h", "7d").
+            offset: Offset to shift window into past (e.g., "1d" = [t-window-1d, t-1d]).
+                Must be a multiple of feature_granularity. Default is "0" (no offset).
+            **params: Additional parameters for the aggregation.
+        """
+        self._function = function
+        self._column = column
+        self._window = window
+        self._offset = offset
+        self._params = params
+        self._alias: Optional[str] = None
+
+    def alias(self, name: str, case_sensitive: bool = False) -> Feature:
+        """Set the output column name for this feature.
+
+        Args:
+            name: The output column name.
+            case_sensitive: If True, preserve the exact case of the name (will be quoted in SQL).
+                If False (default), the name will be converted to uppercase (Snowflake default).
+
+        Returns:
+            Self for method chaining.
+        """
+        # Store with quotes if case-sensitive, uppercase if case-insensitive
+        self._alias = f'"{name}"' if case_sensitive else name.upper()
+        return self
+
+    def to_spec(self) -> AggregationSpec:
+        """Convert to an AggregationSpec for internal use.
+
+        Returns:
+            The AggregationSpec representation.
+        """
+        output_column = self._alias if self._alias else self._default_output_name()
+        return AggregationSpec(
+            function=self._function,
+            source_column=self._column,
+            window=self._window,
+            output_column=output_column,
+            offset=self._offset,
+            params=self._params,
+        )
+
+    def _default_output_name(self) -> str:
+        """Generate a default output column name."""
+        if is_lifetime_window(self._window):
+            window_suffix = "lifetime"
+        else:
+            window_suffix = self._window.replace(" ", "").lower()
+        base_name = f"{self._column}_{self._function.value}_{window_suffix}"
+        return base_name.upper()
+
+    # Factory methods for creating features
+
+    @classmethod
+    def sum(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a SUM aggregation feature.
+
+        Args:
+            column: The column to sum.
+            window: The lookback window (e.g., "24h").
+            offset: Offset to shift window into past (e.g., "7d" = previous week).
+                Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for SUM aggregation.
+
+        Example::
+
+            >>> amount_sum = Feature.sum("amount", "24h").alias("total_amount")
+            >>> prev_week_sum = Feature.sum("amount", "7d", offset="7d").alias("prev_week")
+        """
+        return cls(AggregationType.SUM, column, window, offset)
+
+    @classmethod
+    def count(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a COUNT aggregation feature.
+
+        Args:
+            column: The column to count.
+            window: The lookback window (e.g., "7d").
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for COUNT aggregation.
+
+        Example::
+
+            >>> txn_count = Feature.count("transaction_id", "7d")
+        """
+        return cls(AggregationType.COUNT, column, window, offset)
+
+    @classmethod
+    def avg(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create an AVG aggregation feature.
+
+        Args:
+            column: The column to average.
+            window: The lookback window.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for AVG aggregation.
+
+        Example::
+
+            >>> avg_amount = Feature.avg("amount", "24h")
+        """
+        return cls(AggregationType.AVG, column, window, offset)
+
+    @classmethod
+    def min(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a MIN aggregation feature.
+
+        Args:
+            column: The column to find minimum of.
+            window: The lookback window (e.g., "24h").
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for MIN aggregation.
+
+        Example::
+
+            >>> min_price = Feature.min("price", "24h")
+        """
+        return cls(AggregationType.MIN, column, window, offset)
+
+    @classmethod
+    def max(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a MAX aggregation feature.
+
+        Args:
+            column: The column to find maximum of.
+            window: The lookback window (e.g., "24h").
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for MAX aggregation.
+
+        Example::
+
+            >>> max_price = Feature.max("price", "24h")
+        """
+        return cls(AggregationType.MAX, column, window, offset)
+
+    @classmethod
+    def std(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a STD (standard deviation) aggregation feature.
+
+        Args:
+            column: The column to compute standard deviation for.
+            window: The lookback window.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for STD aggregation.
+
+        Example::
+
+            >>> price_std = Feature.std("price", "24h")
+        """
+        return cls(AggregationType.STD, column, window, offset)
+
+    @classmethod
+    def var(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create a VAR (variance) aggregation feature.
+
+        Args:
+            column: The column to compute variance for.
+            window: The lookback window.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for VAR aggregation.
+
+        Example::
+
+            >>> price_var = Feature.var("price", "24h")
+        """
+        return cls(AggregationType.VAR, column, window, offset)
+
+    @classmethod
+    def last_n(cls, column: str, window: str, *, n: int, offset: str = "0") -> Feature:
+        """Create a LAST_N aggregation feature.
+
+        Collects the N most recent values within the window.
+
+        Args:
+            column: The column to collect values from.
+            window: The lookback window.
+            n: Number of values to collect.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for LAST_N aggregation.
+
+        Example::
+
+            >>> recent_pages = Feature.last_n("page_id", "1h", n=10)
+        """
+        return cls(AggregationType.LAST_N, column, window, offset, n=n)
+
+    @classmethod
+    def last_distinct_n(cls, column: str, window: str, *, n: int, offset: str = "0") -> Feature:
+        """Create a LAST_DISTINCT_N aggregation feature.
+
+        Collects the N most recent distinct values within the window.
+
+        Args:
+            column: The column to collect values from.
+            window: The lookback window.
+            n: Number of distinct values to collect.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for LAST_DISTINCT_N aggregation.
+
+        Example::
+
+            >>> recent_categories = Feature.last_distinct_n("category", "24h", n=5)
+        """
+        return cls(AggregationType.LAST_DISTINCT_N, column, window, offset, n=n)
+
+    @classmethod
+    def first_n(cls, column: str, window: str, *, n: int, offset: str = "0") -> Feature:
+        """Create a FIRST_N aggregation feature.
+
+        Collects the N oldest values within the window.
+
+        Args:
+            column: The column to collect values from.
+            window: The lookback window.
+            n: Number of values to collect.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for FIRST_N aggregation.
+
+        Example::
+
+            >>> first_pages = Feature.first_n("page_id", "1h", n=10)
+        """
+        return cls(AggregationType.FIRST_N, column, window, offset, n=n)
+
+    @classmethod
+    def first_distinct_n(cls, column: str, window: str, *, n: int, offset: str = "0") -> Feature:
+        """Create a FIRST_DISTINCT_N aggregation feature.
+
+        Collects the N oldest distinct values within the window.
+
+        Args:
+            column: The column to collect values from.
+            window: The lookback window.
+            n: Number of distinct values to collect.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for FIRST_DISTINCT_N aggregation.
+
+        Example::
+
+            >>> first_categories = Feature.first_distinct_n("category", "24h", n=5)
+        """
+        return cls(AggregationType.FIRST_DISTINCT_N, column, window, offset, n=n)
+
+    @classmethod
+    def approx_count_distinct(cls, column: str, window: str, offset: str = "0") -> Feature:
+        """Create an APPROX_COUNT_DISTINCT aggregation feature.
+
+        Estimates the number of distinct values using HyperLogLog algorithm.
+        This is approximate but highly efficient for large datasets.
+
+        Args:
+            column: The column to count distinct values.
+            window: The lookback window.
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for APPROX_COUNT_DISTINCT aggregation.
+
+        Example::
+
+            >>> unique_users = Feature.approx_count_distinct("user_id", "24h")
+        """
+        return cls(AggregationType.APPROX_COUNT_DISTINCT, column, window, offset)
+
+    @classmethod
+    def approx_percentile(cls, column: str, window: str, *, percentile: float = 0.5, offset: str = "0") -> Feature:
+        """Create an APPROX_PERCENTILE aggregation feature.
+
+        Estimates the specified percentile using T-Digest algorithm.
+        This is approximate but highly efficient for large datasets.
+
+        Args:
+            column: The column to compute percentile for.
+            window: The lookback window.
+            percentile: The percentile to estimate (0.0 to 1.0). Default is 0.5 (median).
+            offset: Offset to shift window into past. Default is "0" (no offset).
+
+        Returns:
+            A Feature configured for APPROX_PERCENTILE aggregation.
+
+        Example::
+
+            >>> median_amount = Feature.approx_percentile("amount", "24h", percentile=0.5)
+            >>> p95_latency = Feature.approx_percentile("latency", "1h", percentile=0.95)
+        """
+        return cls(AggregationType.APPROX_PERCENTILE, column, window, offset, percentile=percentile)
+
+    def __repr__(self) -> str:
+        alias_str = f", alias='{self._alias}'" if self._alias else ""
+        params_str = f", params={self._params}" if self._params else ""
+        offset_str = f", offset='{self._offset}'" if self._offset != "0" else ""
+        return (
+            f"Feature({self._function.value}, column='{self._column}', "
+            f"window='{self._window}'{offset_str}{params_str}{alias_str})"
+        )