私はここでいくつかの投稿に基づいて評価者クラスを作成しました。基本的にクラスオブジェクトに書き直したevalの例も使用
しました。
import sys
import ast
import operator as op
import abc
import math
class IEvaluator:
__metaclass__ = abc.ABCMeta
@abc.abstractmethod
def eval_expr(cls, expr, subs): # @NoSelf
'''IMPORTANT: this is class method, overload it with @classmethod!
Evaluate an expression given in the expr string.
:param expr: str. String expression.
:param subs: dict. Dictionary with values to substitute.
:returns: Evaluated expression result.
'''
class Evaluator(IEvaluator):
'''Generic evaluator for a string expression. Uses ast and operator
modules. The expr string is parsed with ast resulting in a node tree.
Then the node tree is recursively traversed and evaluated with operations
from the operator module.
:implements: IEvaluator
'''
@classmethod
def _get_op(cls, node):
'''Get the operator corresponding to the node.
:param node: Operator node type with node.op property.
'''
# supported operators
operators = {
ast.Add: op.add,
ast.Sub: op.sub,
ast.Mult: op.mul,
ast.Div: op.truediv,
ast.Pow: op.pow,
ast.BitXor: op.xor,
ast.USub: op.neg
}
return operators[type(node.op)]
@classmethod
def _get_op_fun(cls, node):
# fun_call = {'sin': math.sin, 'cos': math.cos}[node.func.id]
fun_call = getattr(math, node.func.id)
return fun_call
@classmethod
def _num_op(cls, node, subs):
'''Return the value of the node.
:param node: Value node type with node.n property.
'''
return node.n
@classmethod
def _bin_op(cls, node, subs):
'''Eval the left and right nodes, and call the binary operator.
:param node: Binary operator with node.op, node.left, and node.right
properties.
'''
op = cls._get_op(node)
left_node = cls.eval(node.left, subs)
right_node = cls.eval(node.right, subs)
return op(left_node, right_node)
@classmethod
def _unary_op(cls, node, subs):
'''Eval the node operand and call the unary operator.
:param node: Unary operator with node.op and node.operand properties.
'''
op = cls._get_op(node)
return op(cls.eval(node.operand, subs))
@classmethod
def _subs_op(cls, node, subs):
'''Return the value of the variable represented by the node.
:param node: Name node with node.id property to identify the variable.
'''
try:
return subs[node.id]
except KeyError:
raise TypeError(node)
@classmethod
def _call_op(cls, node, subs):
arg_list = []
for node_arg in node.args:
arg_list.append(cls.eval(node_arg, subs))
fun_call = cls._get_op_fun(node)
return fun_call(*arg_list)
@classmethod
def eval(cls, node, subs):
'''The node is actually a tree. The node type i.e. type(node) is:
ast.Num, ast.BinOp, ast.UnaryOp or ast.Name.
Depending on the node type the node will have the following properties:
node.n - Nodes value.
node.id - Node id corresponding to a key in the subs dictionary.
node.op - operation node. Type of node.op identifies the operation.
type(node.op) is one of ast.Add, ast.Sub, ast.Mult, ast.Div,
ast.Pow, ast.BitXor, or ast.USub.
node.left or node.right - Binary operation node needs to have links
to left and right nodes.
node.operand - Unary operation node needs to have an operand.
The binary and unary operations call eval recursively.
'''
# The functional logic is:
# if isinstance(node, ast.Num): # <number>
# return node.n
# elif isinstance(node, ast.BinOp): # <left> <operator> <right>
# return operators[type(node.op)](eval_(node.left, subs),
# eval_(node.right, subs))
# elif isinstance(node, ast.UnaryOp): # <operator> <operand> e.g., -1
# return operators[type(node.op)](eval_(node.operand, subs))
# else:
# try:
# return subs[node.id]
# except KeyError:
# raise TypeError(node)
node_type = type(node)
return {
# Value in the expression. Leaf.
ast.Num: cls._num_op, # <number>
# Bin operation with two operands.
ast.BinOp: cls._bin_op, # <left> <operator> <right>
# Unary operation such as neg.
ast.UnaryOp: cls._unary_op, # <operator> <operand> e.g., -1
# Sub the value for the variable. Leaf.
ast.Name: cls._subs_op, # <variable>
ast.Call: cls._call_op
}[node_type](node, subs)
@classmethod
def eval_expr(cls, expr, subs=None):
'''Evaluates a string expression. The expr string is parsed with ast
resulting in a node tree. Then the eval method is used to recursively
traverse and evaluate the nodes. Symbolic params are taken from subs.
:Example:
>>> eval_expr('2^6')
4
>>> eval_expr('2**6')
64
>>> eval_expr('1 + 2*3**(4^5) / (6 + -7)')
-5.0
>>> eval_expr('x + y', {'x': 1, 'y': 2})
3
:param expr: str. String expression.
:param subs: dict. (default: globals of current and calling stack.)
:returns: Result of running the evaluator.
:implements: IEvaluator.eval_expr
'''
# ref: https://stackoverflow.com/a/9558001/3457624
if subs is None:
# Get the globals
frame = sys._getframe()
subs = {}
subs.update(frame.f_globals)
if frame.f_back:
subs.update(frame.f_back.f_globals)
expr_tree = ast.parse(expr, mode='eval').body
return cls.eval(expr_tree, subs)
ここではいくつかの例を示します。
import sympy
from eval_sympy import Evaluator
# test case...
x = sympy.Symbol('x')
y = sympy.Symbol('y')
expr = x * 2 - y ** 2
# z = expr.subs({x:1, y:2})
str_expr = str(expr)
print str_expr
x = 1
y = 2
out0 = Evaluator.eval_expr(str_expr)
print '(x, y): ({}, {})'.format(x, y)
print str_expr, ' = ', out0
subs1 = {'x': 1, 'y': 2}
out1 = Evaluator.eval_expr(str_expr, subs1)
print 'subs: ', subs1
print str_expr, ' = ', out1
sin_subs = {'x': 1, 'y': 2}
sin_out = Evaluator.eval_expr('sin(log10(x*y))', sin_subs)
print 'sin_subs: ', sin_subs
print 'sin(log10(x*y)) = ', sin_out
結果
2*x - y**2
(x, y): (1, 2)
2*x - y**2 = -2
subs: {'y': 2, 'x': 1}
2*x - y**2 = -2
sin_subs: {'y': 2, 'x': 1}
sin(log10(x*y)) = 0.296504042171