def evaluate_expression(expression_string, grids, symbol):
"""Evaluates expression.
Args:
expression_string: String. The univariate expression, for example
'x * x + 1 / x'.
grids: Numpy array with shape [num_grid_points], the points to evaluate
expression.
symbol: String. Symbol of variable in expression.
Returns:
Numpy array with shape [num_grid_points].
"""
try:
expression_on_grids = evaluators.numpy_array_eval(
str(sympy.simplify(expression_string)), arguments={symbol: grids})
except SyntaxError as error:
# NOTE(leeley): In some rare cases, after sympy.simplify(),
# expression_string will contain symbols which can not be parsed,
# for example 'zoo'. If this occurs, evaluate expression without
# simplification.
logging.warning(error)
logging.warning(
'SyntaxError occurs after sympy.simplify(), '
'evaluate %s directly without simplification.', expression_string)
expression_on_grids = evaluators.numpy_array_eval(
expression_string, arguments={symbol: grids})
if np.asarray(expression_on_grids).size == 1:
expression_on_grids = expression_on_grids * np.ones_like(grids)
return expression_on_grids
def test_number(self):
# Integers.
self.assertEqual(evaluators.numpy_array_eval('42'), 42)
self.assertEqual(evaluators.numpy_array_eval('-42'), -42)
# Floats.
self.assertAlmostEqual(evaluators.numpy_array_eval('4.2'), 4.2)
self.assertAlmostEqual(evaluators.numpy_array_eval('-4.2'), -4.2)
def test_malformed_string(self, string):
with self.assertRaisesRegex(SyntaxError, 'Malformed string'):
evaluators.numpy_array_eval(string)
def test_unknown_callable(self):
with self.assertRaisesRegex(SyntaxError, 'Unknown callable: \'foo\''):
evaluators.numpy_array_eval('foo( 4 )', callables={'bar': np.sin})
def test_unknown_argument(self):
with self.assertRaisesRegex(SyntaxError, 'Unknown argument: \'x\''):
evaluators.numpy_array_eval('x', arguments={'y': 2})
def test_callables_not_dict(self):
with self.assertRaisesRegex(ValueError,
'Input callables expected to be a dict'):
evaluators.numpy_array_eval('sqrt( 4 )', callables=[np.sqrt])
def test_arguments_not_dict(self):
with self.assertRaisesRegex(ValueError,
'Input arguments expected to be a dict'):
evaluators.numpy_array_eval('x', arguments=[42])
def test_one_zero_element_in_divisor_array_expression(self):
arguments = {'x': 1., 'y': np.array([2., 0., 0.5])}
np.testing.assert_allclose(
evaluators.numpy_array_eval('x / y', arguments=arguments),
np.array([0.5, 0., 2.]))
def test_callables(self, string, arguments, expected):
np.testing.assert_allclose(
evaluators.numpy_array_eval(string, arguments=arguments), expected)
def test_numpy_array_and_number_binary_operators(self, string, expected):
arguments = {'x': np.array([1., 2., 3.]), 'y': 10.}
np.testing.assert_allclose(
evaluators.numpy_array_eval(string, arguments=arguments), expected)
def test_number_binary_operators_power(self, arguments, expected):
self.assertAlmostEqual(
evaluators.numpy_array_eval('x ** y', arguments=arguments),
expected)
def test_number_binary_operators(self, string, expected):
arguments = {'x': 1., 'y': 2.}
self.assertAlmostEqual(
evaluators.numpy_array_eval(string, arguments=arguments), expected)
def test_unary_operator(self, arguments, expected):
np.testing.assert_allclose(
evaluators.numpy_array_eval('-x', arguments=arguments), expected)
