input_parser.py

import ast
import re
import numpy as np

MAX_INPUT_TEXT_LEN = 1000
MAX_TOTAL_ELEMENTS = 50000
MAX_ARRAY_RANK = 2
MAX_NUMERIC_MAGNITUDE = 1e9

def split_top_level_assignments(text):
    """
    Split an assignment string (e.g. 'W = [[1, 2], [3, 4]], x = [1, 1], b = 0.5')
    only at commas that are at bracket depth 0.
    """
    chunks = []
    current = []
    depth = 0
    for char in text:
        if char in '([{':
            depth += 1
            current.append(char)
        elif char in ')]}':
            depth -= 1
            if depth < 0:
                raise ValueError("Unmatched closing bracket in variable values.")
            current.append(char)
        elif char == ',' and depth == 0:
            chunks.append(''.join(current).strip())
            current = []
        else:
            current.append(char)
    if depth != 0:
        raise ValueError("Unmatched opening bracket in variable values.")
    last = ''.join(current).strip()
    if last:
        chunks.append(last)
    return chunks


def _validate_and_convert_literal(node):
    """
    Recursively inspect an AST literal node and convert to a Python float or nested list of floats.
    Strictly forbids function calls, comprehensions, variables, strings, booleans, etc.
    """
    if isinstance(node, ast.Constant):
        # In Python 3.8+, booleans are an instance of int/Constant with value bool
        if isinstance(node.value, bool):
            raise ValueError("Boolean literals are not allowed.")
        if not isinstance(node.value, (int, float)):
            raise ValueError(f"Only numeric literals are allowed, got {type(node.value).__name__}.")
        val = float(node.value)
        if np.isnan(val) or np.isinf(val):
            raise ValueError("Values must be finite real numbers.")
        if abs(val) > MAX_NUMERIC_MAGNITUDE:
            raise ValueError(f"Numeric literal exceeds maximum allowed magnitude ({MAX_NUMERIC_MAGNITUDE}).")
        return val

    if hasattr(ast, "Num") and isinstance(node, ast.Num):
        val = float(node.n)
        if abs(val) > MAX_NUMERIC_MAGNITUDE:
            raise ValueError(f"Numeric literal exceeds maximum allowed magnitude.")
        return val

    if isinstance(node, ast.UnaryOp):
        if isinstance(node.op, ast.UAdd):
            return +_validate_and_convert_literal(node.operand)
        elif isinstance(node.op, ast.USub):
            return -_validate_and_convert_literal(node.operand)
        raise ValueError(f"Unsupported unary operator in literal: {type(node.op).__name__}")

    if isinstance(node, ast.List):
        if len(node.elts) == 0:
            raise ValueError("Empty lists or arrays are not allowed.")
        return [_validate_and_convert_literal(elt) for elt in node.elts]

    raise ValueError(f"Disallowed construct in literal: {type(node).__name__}")


def parse_literal_value(val_str):
    """
    Safely parse a literal string (number, vector list, or matrix list) into a float or np.ndarray.
    Enforces rectangular matrix shape, finite elements, rank <= 2, and element count budget.
    """
    val_str = val_str.strip()
    if not val_str:
        raise ValueError("Empty value literal.")

    try:
        parsed = ast.parse(val_str, mode="eval")
    except SyntaxError as e:
        raise ValueError(f"Syntax error in literal '{val_str}': {e.msg}")

    val = _validate_and_convert_literal(parsed.body)

    if isinstance(val, (int, float)):
        return float(val)

    # Convert nested lists to numpy array
    try:
        arr = np.array(val, dtype=np.float64)
    except ValueError as e:
        raise ValueError(f"Malformed or ragged array structure: {e}")

    if arr.ndim > MAX_ARRAY_RANK:
        raise ValueError(f"Array rank {arr.ndim} exceeds maximum supported rank ({MAX_ARRAY_RANK}).")

    if arr.size == 0:
        raise ValueError("Empty array dimensions are not allowed.")

    if arr.size > MAX_TOTAL_ELEMENTS:
        raise ValueError(f"Array size ({arr.size} elements) exceeds limit ({MAX_TOTAL_ELEMENTS}).")

    if not np.all(np.isfinite(arr)):
        raise ValueError("Array contains nonfinite numbers (NaN or Inf).")

    return arr


def parse_variable_assignments(text):
    """
    Parse a comma-separated string of variable assignments into a dictionary:
        { "W": np.ndarray, "x": np.ndarray, "b": float, ... }
    Validates variable names, duplicate keys, and bounds.
    """
    if text is None or not text.strip():
        return {}

    cleaned = text.strip()
    if len(cleaned) > MAX_INPUT_TEXT_LEN:
        raise ValueError(f"Input text exceeds maximum length ({MAX_INPUT_TEXT_LEN} characters).")

    chunks = split_top_level_assignments(cleaned)
    feed_dict = {}
    total_elements = 0

    for chunk in chunks:
        if not chunk:
            continue
        if "=" not in chunk:
            raise ValueError(f"Expected assignment 'var = value', got '{chunk}'")

        var_part, val_part = chunk.split("=", 1)
        var_name = var_part.strip()
        val_str = val_part.strip()

        if not re.match(r"^[a-zA-Z][a-zA-Z0-9_]*$", var_name):
            raise ValueError(f"Invalid variable identifier: '{var_name}'")

        if var_name in feed_dict:
            raise ValueError(f"Duplicate assignment for variable '{var_name}'")

        parsed_val = parse_literal_value(val_str)
        if isinstance(parsed_val, np.ndarray):
            total_elements += parsed_val.size
        else:
            total_elements += 1

        if total_elements > MAX_TOTAL_ELEMENTS:
            raise ValueError(f"Total element count exceeds limit ({MAX_TOTAL_ELEMENTS}).")

        feed_dict[var_name] = parsed_val

    return feed_dict