def test_option():
assert_dshape_equal(Option(int32), Option(int32))
with pytest.raises(AssertionError) as e:
assert_dshape_equal(Option(int32), Option(float32))
assert "'int32' != 'float32'" in str(e.value)
assert '_.ty' in str(e.value)
def test_unite_records():
assert (discover([{
'name': 'Alice',
'balance': 100
}, {
'name': 'Bob',
'balance': ''
}]) == 2 * Record([['balance', Option(int64)], ['name', string]]))
assert (discover([{
'name': 'Alice',
's': 'foo'
}, {
'name': 'Bob',
's': None
}]) == 2 * Record([['name', string], ['s', Option(string)]]))
assert (discover([{
'name': 'Alice',
's': 'foo',
'f': 1.0
}, {
'name': 'Bob',
's': None,
'f': None
}]) == 2 * Record([['f', Option(float64)], ['name', string],
['s', Option(string)]]))
def test_unite_tuples():
assert (discover([[1, 1, 'hello'], [1, '', ''], [1, 1, 'hello']]) == 3 *
Tuple([int64, Option(int64), Option(string)]))
assert (discover([[1, 1, 'hello', 1], [1, '', '', 1], [1, 1, 'hello',
1]]) == 3 *
Tuple([int64, Option(int64),
Option(string), int64]))
def build_datashape(sample_data):
""" Build a datashape from sample_data
In case of multiple dimensions, the firt is replace by 'var' to
handle varing time samples in models
"""
if isinstance(sample_data, collections.Mapping):
return Record((k,build_datashape(v)) for k,v in sample_data.items() )
else:
ds = discover(sample_data)
if len(ds) > 1:
ds = var * ds.subarray(1)
else:
ds = Option( ds )
return ds
def test_unite_missing_values():
assert unite([int32, null, int32]) == 3 * Option(int32)
assert unite([string, null, int32])
def test_to_numpy_fails():
ds = var * int32
with pytest.raises(TypeError):
to_numpy(ds)
with pytest.raises(TypeError):
to_numpy(Option(int32))
def test_tuple_str():
assert str(Tuple([Option(int32), float64])) == '(?int32, float64)'
def test_funcproto_attrs():
f = dshape('(int32, ?float64) -> {a: ?string}').measure
assert f.restype == DataShape(Record([('a', Option(String()))]))
assert f.argtypes == (DataShape(int32), DataShape(Option(float64)))
def test_option_to_numpy_dtype_fails(base):
Option(base).to_numpy_dtype()
def test_option_to_numpy_dtype(base, expected):
assert Option(base).to_numpy_dtype() == np.dtype(expected)
def test_option_timedelta_to_numpy(unit):
assert (Option(TimeDelta(unit=unit)).to_numpy_dtype() == np.dtype(
'timedelta64[%s]' % unit))
def test_option_datetime_to_numpy():
assert Option(DateTime()).to_numpy_dtype() == np.dtype('datetime64[us]')
def test_option_sanitizes_strings():
assert Option('float32').ty == dshape('float32').measure
))
def test_invalid_record_literal(invalid):
assert pytest.raises(TypeError, getitem, R, invalid)
@pytest.mark.parametrize(['names', 'typ'],
[(['foo', b'\xc4\x87'.decode('utf8')], unicode),
(['foo', 'bar'], str), (list(u'ab'), unicode)])
def test_unicode_record_names(names, typ):
types = [int64, float64]
record = Record(list(zip(names, types)))
string_type, = set(map(type, record.names))
assert record.names == names
assert record.types == types
assert all(isinstance(s, string_type) for s in record.names)
equiv_dshape_pairs = [(dshape('?string'), Option('string')),
(dshape('string'), string),
(dshape('datetime'), datetime_),
(dshape('?datetime'), Option(datetime_)),
(dshape('10 * int32'), 10 * int32),
(dshape('var * ?int32'), var * Option(int32)),
(dshape('10 * ?float64'), Fixed(10) * Option(float64))]
@pytest.mark.parametrize('a,b', equiv_dshape_pairs)
def test_hash_and_eq_consistency(a, b):
assert a == b
assert hash(a) == hash(b)
