src.asqi.metric_expression

Safe metric expression evaluator for score card indicators.

Supports arithmetic operations and aggregation functions while maintaining security by using AST parsing instead of eval().

Attributes

logger

Classes

MetricExpressionEvaluator

Safe evaluator for metric expressions in score cards.

Module Contents

src.asqi.metric_expression.logger

class src.asqi.metric_expression.MetricExpressionEvaluator

Safe evaluator for metric expressions in score cards.

Supports: - Arithmetic operators: +, -, *, / - Aggregation functions: min(), max(), avg() - Numeric literals and parentheses - Complex formulas: ‘0.7 * accuracy + 0.3 * relevance’

Does NOT support: - Code execution (no eval/exec) - Arbitrary function calls - Variable assignment - Imports or other Python statements

Examples:

>>> evaluator = MetricExpressionEvaluator()
>>> # Simple variable
>>> evaluator.evaluate_expression("accuracy", {"accuracy": 0.85})
0.85
>>> # Weighted average
>>> evaluator.evaluate_expression("0.7 * a + 0.3 * b", {"a": 0.8, "b": 0.9})
0.83
>>> # Min function
>>> evaluator.evaluate_expression("min(x, y, z)", {"x": 0.9, "y": 0.7, "z": 0.8})
0.7

ALLOWED_OPS

ALLOWED_FUNCTIONS

parse_expression(expression: str) → ast.Expression

Parse an expression string into an AST.

Args:

expression: The expression string to parse

Returns:

Parsed AST expression

Raises:

MetricExpressionError: If parsing fails or contains unsafe operations

evaluate_expression(expression: str, metric_values: Dict[str, int | float]) → int | float

Evaluate a metric expression with provided metric values.

Args:

expression: The expression string to evaluate metric_values: Dictionary mapping metric paths to their numeric values

Returns:

The computed numeric result

Raises:

MetricExpressionError: If evaluation fails

Example:

>>> evaluator.evaluate_expression(
...     "0.5 * accuracy + 0.5 * relevance",
...     {"accuracy": 0.8, "relevance": 0.9}
... )
0.85