def test_numpy_float_types():
e = NumpyJSONEncoder()
numpy_floats = (np.float, np.float_, np.float16, np.float32, np.float64)
for float_type in numpy_floats:
assert e.encode(float_type(2.5)) == '2.5'
def test_complex_types():
e = NumpyJSONEncoder()
assert e.encode(complex(1, 2)) == \
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}'
for complex_type in numpy_complex:
assert e.encode(complex_type(complex(1, 2))) == \
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}'
def test_complex_types():
e = NumpyJSONEncoder()
assert e.encode(complex(1, 2)) == \
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}'
assert e.encode(np.complex(1, 2)) == \
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}'
assert e.encode(np.complex64(complex(1, 2))) == \
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}'
def test_numpy_float_types(self):
e = NumpyJSONEncoder()
numpy_floats = (np.float, np.float_, np.float16, np.float32,
np.float64)
for float_type in numpy_floats:
self.assertEqual(e.encode(float_type(2.5)), '2.5')
def test_complex_types(self):
e = NumpyJSONEncoder()
self.assertEqual(e.encode(complex(1, 2)),
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}')
self.assertEqual(e.encode(np.complex(1, 2)),
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}')
self.assertEqual(e.encode(np.complex64(complex(1, 2))),
'{"__dtype__": "complex", "re": 1.0, "im": 2.0}')
def test_numpy_array(self):
e = NumpyJSONEncoder()
self.assertEqual(e.encode(np.array([1, 0, 0])), '[1, 0, 0]')
self.assertEqual(e.encode(np.arange(1.0, 3.0, 1.0)), '[1.0, 2.0]')
self.assertEqual(e.encode(np.meshgrid((1, 2), (3, 4))),
'[[[1, 2], [1, 2]], [[3, 3], [4, 4]]]')
def test_numpy_int_types(self):
e = NumpyJSONEncoder()
numpy_ints = (np.int, np.int_, np.int8, np.int16, np.int32, np.int64,
np.intc, np.intp, np.uint, np.uint8, np.uint16,
np.uint32, np.uint64, np.uintc, np.uintp)
for int_type in numpy_ints:
self.assertEqual(e.encode(int_type(3)), '3')
def test_numpy_int_types():
e = NumpyJSONEncoder()
numpy_ints = (np.int, np.int_, np.int8, np.int16, np.int32, np.int64,
np.intc, np.intp, np.uint, np.uint8, np.uint16, np.uint32,
np.uint64, np.uintc, np.uintp)
for int_type in numpy_ints:
assert e.encode(int_type(3)) == '3'
def test_numpy_array():
e = NumpyJSONEncoder()
assert e.encode(np.array([1, 0, 0])) == \
'[1, 0, 0]'
assert e.encode(np.arange(1.0, 3.0, 1.0)) == \
'[1.0, 2.0]'
assert e.encode(np.meshgrid((1, 2), (3, 4))) == \
'[[[1, 2], [1, 2]], [[3, 3], [4, 4]]]'
def test_non_serializable():
"""
Test that non-serializable objects are serialzed to their
string representation
"""
e = NumpyJSONEncoder()
class Dummy:
def __str__(self):
return 'i am a dummy with \\ slashes /'
assert e.encode(Dummy()) == \
'"i am a dummy with \\\\ slashes /"'
def test_python_types():
e = NumpyJSONEncoder()
# test basic python types
od = OrderedDict()
od['a'] = 0
od['b'] = 1
testinput = [None, True, False, 10, float(10.), 'hello', od]
testoutput = [
'null', 'true', 'false', '10', '10.0', '"hello"', '{"a": 0, "b": 1}'
]
for d, r in zip(testinput, testoutput):
v = e.encode(d)
if isinstance(d, dict):
assert v == r
else:
assert v == r
def test_python_types(self):
e = NumpyJSONEncoder()
# test basic python types
od = OrderedDict()
od['a'] = 0
od['b'] = 1
testinput = [None, True, False, 10, float(10.), 'hello',
od]
testoutput = ['null', 'true', 'false', '10', '10.0', '"hello"',
'{"a": 0, "b": 1}']
for d, r in zip(testinput, testoutput):
v = e.encode(d)
if type(d) == dict:
self.assertDictEqual(v, r)
else:
self.assertEqual(v, r)
def test_object_with_serialization_method():
"""
Test that objects with `_JSONEncoder` method are serialized via
calling that method
"""
e = NumpyJSONEncoder()
class Dummy:
def __init__(self):
self.confession = 'a_dict_addict'
def __str__(self):
return 'me as a string'
def _JSONEncoder(self):
return {'i_am_actually': self.confession}
assert e.encode(Dummy()) == \
'{"i_am_actually": "a_dict_addict"}'
def testNumpyJSONEncoder(self):
e = NumpyJSONEncoder()
# test basic python types
od = OrderedDict()
od['a'] = 0
od['b'] = 1
testinput = [10, float(10.), 'hello', od]
testoutput = ['10', '10.0', '"hello"', '{"a": 0, "b": 1}']
# int
for d, r in zip(testinput, testoutput):
v = e.encode(d)
if type(d) == dict:
self.assertDictEqual(v, r)
else:
self.assertEqual(v, r)
# test numpy array
x = np.array([1, 0, 0])
v = e.encode(x)
self.assertEqual(v, '[1, 0, 0]')
# test class
class dummy(object):
pass
# test that does not raise, do not care about
# return value
e.encode(dummy())
def test_numpy_bool_type(self):
e = NumpyJSONEncoder()
self.assertEqual(e.encode(np.bool_(True)), 'true')
self.assertEqual(e.encode(np.int8(5) == 5), 'true')
self.assertEqual(e.encode(np.array([8, 5]) == 5), '[false, true]')
def test_UFloat_type():
e = NumpyJSONEncoder()
assert e.encode(uncertainties.ufloat(1.0, 2.0)) == \
'{"__dtype__": "UFloat", "nominal_value": 1.0, "std_dev": 2.0}'
def test_numpy_bool_type():
e = NumpyJSONEncoder()
assert e.encode(np.bool_(True)) == 'true'
assert e.encode(np.int8(5) == 5) == 'true'
assert e.encode(np.array([8, 5]) == 5) == '[false, true]'
def test_numpy_float_types():
e = NumpyJSONEncoder()
for float_type in numpy_floats:
assert e.encode(float_type(2.5)) == '2.5'
def test_numpy_int_types():
e = NumpyJSONEncoder()
for int_type in numpy_ints:
assert e.encode(int_type(3)) == '3'
