sqlspec 0.16.1__cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl

sqlspec/core/hashing.py

@@ -0,0 +1,310 @@
+"""Statement hashing utilities for cache key generation.
+
+This module provides centralized hashing logic for SQL statements,
+including expressions, parameters, filters, and complete SQL objects.
+
+Also supports fine-grained AST sub-expression caching.
+"""
+
+from typing import TYPE_CHECKING, Any, Optional
+
+from sqlglot import exp
+
+from sqlspec.utils.type_guards import is_typed_parameter
+
+if TYPE_CHECKING:
+    from sqlspec.core.filters import StatementFilter
+    from sqlspec.core.statement import SQL
+
+__all__ = (
+    "hash_expression",
+    "hash_expression_node",
+    "hash_optimized_expression",
+    "hash_parameters",
+    "hash_sql_statement",
+)
+
+
+def hash_expression(expr: Optional[exp.Expression], _seen: Optional[set[int]] = None) -> int:
+    """Generate deterministic hash from AST structure.
+
+    Args:
+        expr: SQLGlot Expression to hash
+        _seen: Set of seen object IDs to handle circular references
+
+    Returns:
+        Deterministic hash of the AST structure
+    """
+    if expr is None:
+        return hash(None)
+
+    if _seen is None:
+        _seen = set()
+
+    expr_id = id(expr)
+    if expr_id in _seen:
+        return hash(expr_id)
+
+    _seen.add(expr_id)
+
+    # Build hash from type and args
+    components: list[Any] = [type(expr).__name__]
+
+    for key, value in sorted(expr.args.items()):
+        components.extend((key, _hash_value(value, _seen)))
+
+    return hash(tuple(components))
+
+
+def _hash_value(value: Any, _seen: set[int]) -> int:
+    """Hash different value types consistently.
+
+    Args:
+        value: Value to hash (can be Expression, list, dict, or primitive)
+        _seen: Set of seen object IDs to handle circular references
+
+    Returns:
+        Deterministic hash of the value
+    """
+    if isinstance(value, exp.Expression):
+        return hash_expression(value, _seen)
+    if isinstance(value, list):
+        return hash(tuple(_hash_value(v, _seen) for v in value))
+    if isinstance(value, dict):
+        items = sorted((k, _hash_value(v, _seen)) for k, v in value.items())
+        return hash(tuple(items))
+    if isinstance(value, tuple):
+        return hash(tuple(_hash_value(v, _seen) for v in value))
+    # Primitives: str, int, bool, None, etc.
+    return hash(value)
+
+
+def hash_parameters(
+    positional_parameters: Optional[list[Any]] = None,
+    named_parameters: Optional[dict[str, Any]] = None,
+    original_parameters: Optional[Any] = None,
+) -> int:
+    """Generate hash for SQL parameters.
+
+    Args:
+        positional_parameters: List of positional parameters
+        named_parameters: Dictionary of named parameters
+        original_parameters: Original parameters (for execute_many)
+
+    Returns:
+        Combined hash of all parameters
+    """
+    param_hash = 0
+
+    # Hash positional parameters
+    if positional_parameters:
+        from sqlspec.core.parameters import TypedParameter
+
+        hashable_parameters: list[tuple[Any, Any]] = []
+        for param in positional_parameters:
+            if isinstance(param, TypedParameter):
+                if isinstance(param.value, (list, dict)):
+                    hashable_parameters.append((repr(param.value), param.original_type))
+                else:
+                    hashable_parameters.append((param.value, param.original_type))
+            elif isinstance(param, (list, dict)):
+                # Convert unhashable types to hashable representations
+                hashable_parameters.append((repr(param), "unhashable"))
+            else:
+                # Check if param itself contains unhashable types
+                try:
+                    hash(param)
+                    hashable_parameters.append((param, "primitive"))
+                except TypeError:
+                    # If unhashable, convert to string representation
+                    hashable_parameters.append((repr(param), "unhashable_repr"))
+
+        param_hash ^= hash(tuple(hashable_parameters))
+
+    if named_parameters:
+        # Handle unhashable types in named parameters
+        hashable_items = []
+        for key, value in sorted(named_parameters.items()):
+            if is_typed_parameter(value):
+                # For TypedParameter, hash its value with type info
+                if isinstance(value.value, (list, dict)):
+                    hashable_items.append((key, (repr(value.value), value.original_type)))
+                else:
+                    hashable_items.append((key, (value.value, value.original_type)))
+            elif isinstance(value, (list, dict)):
+                hashable_items.append((key, (repr(value), "unhashable")))
+            else:
+                hashable_items.append((key, (value, "primitive")))
+        param_hash ^= hash(tuple(hashable_items))
+
+    # Hash original parameters (important for execute_many)
+    if original_parameters is not None:
+        if isinstance(original_parameters, list):
+            # For execute_many, hash the count and first few items to avoid
+            param_hash ^= hash(("original_count", len(original_parameters)))
+            if original_parameters:
+                # Hash first 3 items as representatives
+                sample_size = min(3, len(original_parameters))
+                sample_hash = hash(repr(original_parameters[:sample_size]))
+                param_hash ^= hash(("original_sample", sample_hash))
+        else:
+            param_hash ^= hash(("original", repr(original_parameters)))
+
+    return param_hash
+
+
+def _hash_filter_value(value: Any) -> int:
+    try:
+        return hash(value)
+    except TypeError:
+        return hash(repr(value))
+
+
+def hash_filters(filters: Optional[list["StatementFilter"]] = None) -> int:
+    """Generate hash for statement filters.
+
+    Args:
+        filters: List of statement filters
+
+    Returns:
+        Hash of the filters
+    """
+    if not filters:
+        return 0
+
+    # Use class names and any hashable attributes
+    filter_components = []
+    for f in filters:
+        # Use class name as primary identifier
+        components: list[Any] = [f.__class__.__name__]
+
+        # Add any hashable attributes if available
+        # Use getattr with default instead of hasattr for mypyc compatibility
+        filter_dict = getattr(f, "__dict__", None)
+        if filter_dict is not None:
+            for key, value in sorted(filter_dict.items()):
+                components.append((key, _hash_filter_value(value)))
+
+        filter_components.append(tuple(components))
+
+    return hash(tuple(filter_components))
+
+
+def hash_sql_statement(statement: "SQL") -> str:
+    """Generate a complete cache key for a SQL statement.
+
+    This centralizes all the complex hashing logic that was previously
+    scattered across different parts of the codebase.
+
+    Args:
+        statement: SQL statement object
+
+    Returns:
+        Cache key string
+    """
+    from sqlspec.utils.type_guards import is_expression
+
+    # Hash the expression or raw SQL
+    if is_expression(statement._statement):
+        expr_hash = hash_expression(statement._statement)
+    else:
+        expr_hash = hash(statement._raw_sql)
+
+    # Hash all parameters
+    param_hash = hash_parameters(
+        positional_parameters=statement._positional_parameters,
+        named_parameters=statement._named_parameters,
+        original_parameters=statement._original_parameters,
+    )
+
+    # Hash filters
+    filter_hash = hash_filters(statement._filters)
+
+    # Combine with other state
+    state_components = [
+        expr_hash,
+        param_hash,
+        filter_hash,
+        hash(statement._dialect),
+        hash(statement._is_many),
+        hash(statement._is_script),
+    ]
+
+    return f"sql:{hash(tuple(state_components))}"
+
+
+def hash_expression_node(node: exp.Expression, include_children: bool = True, dialect: Optional[str] = None) -> str:
+    """Generate a cache key for an individual expression node.
+
+    This is used for sub-expression caching where we want to cache
+    frequently used SQL fragments like complex JOIN clauses or WHERE conditions.
+
+    Args:
+        node: The expression node to hash
+        include_children: Whether to include child nodes in the hash
+        dialect: SQL dialect for context-aware hashing
+
+    Returns:
+        Cache key string for the expression node
+    """
+    if include_children:
+        # Full structural hash including all children
+        node_hash = hash_expression(node)
+    else:
+        # Shallow hash - just the node type and immediate args
+        components: list[Any] = [type(node).__name__]
+        for key, value in sorted(node.args.items()):
+            if not isinstance(value, (list, exp.Expression)):
+                # Only include primitive values, not child expressions
+                components.extend((key, hash(value)))
+        node_hash = hash(tuple(components))
+
+    dialect_part = f":{dialect}" if dialect else ""
+    return f"expr{dialect_part}:{node_hash}"
+
+
+def hash_optimized_expression(
+    expr: exp.Expression,
+    dialect: str,
+    schema: Optional[dict[str, Any]] = None,
+    optimizer_settings: Optional[dict[str, Any]] = None,
+) -> str:
+    """Generate a cache key for optimized expressions.
+
+    This creates a unique key that captures the expression structure,
+    dialect, schema context, and optimizer settings to ensure we only
+    reuse optimized expressions when all context matches.
+
+    Args:
+        expr: The unoptimized expression
+        dialect: Target SQL dialect
+        schema: Schema information
+        optimizer_settings: Additional optimizer configuration
+
+    Returns:
+        Cache key string for the optimized expression
+    """
+    # Base expression hash
+    expr_hash = hash_expression(expr)
+
+    # Schema hash - simplified representation
+    schema_hash = 0
+    if schema:
+        # Hash table names and column counts to avoid deep schema hashing
+        schema_items = []
+        for table_name, table_schema in sorted(schema.items()):
+            if isinstance(table_schema, dict):
+                schema_items.append((table_name, len(table_schema)))
+            else:
+                schema_items.append((table_name, hash("unknown")))
+        schema_hash = hash(tuple(schema_items))
+
+    # Optimizer settings hash
+    settings_hash = 0
+    if optimizer_settings:
+        settings_items = sorted(optimizer_settings.items())
+        settings_hash = hash(tuple(settings_items))
+
+    # Combine all components
+    components = (expr_hash, dialect, schema_hash, settings_hash)
+    return f"opt:{hash(components)}"