def write_reread(t: Tafra) -> None:
t.to_csv(write_path)
t2 = Tafra.read_csv(write_path, dtypes=t.dtypes)
for c1, c2 in zip(t.columns, t2.columns):
assert np.array_equal(t.data[c1], t2.data[c2])
assert np.array_equal(t.dtypes[c1], t2.dtypes[c2])
def test_csv() -> None:
write_path = 'test/test_to_csv.csv'
def write_reread(t: Tafra) -> None:
t.to_csv(write_path)
t2 = Tafra.read_csv(write_path, dtypes=t.dtypes)
for c1, c2 in zip(t.columns, t2.columns):
assert np.array_equal(t.data[c1], t2.data[c2])
assert np.array_equal(t.dtypes[c1], t2.dtypes[c2])
# straightforward CSV - inference heuristic works
path = Path('test/ex1.csv')
t = Tafra.read_csv(path)
assert t.dtypes['a'] == 'int32'
assert t.dtypes['b'] == 'bool'
assert t.dtypes['c'] == 'float64'
assert t.rows == 6
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)
# test again with TextIOWrapper
with open('test/ex1.csv', 'r') as f:
t = Tafra.read_csv(f)
assert t.dtypes['a'] == 'int32'
assert t.dtypes['b'] == 'bool'
assert t.dtypes['c'] == 'float64'
assert t.rows == 6
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)
with open(write_path, 'w') as f:
t.to_csv(f)
with pytest.raises(ValueError) as e:
with open(write_path) as f:
t.to_csv(f)
# short CSV - ends during inference period
t = Tafra.read_csv('test/ex2.csv')
assert t.dtypes['a'] == 'int32'
assert t.dtypes['b'] == 'bool'
assert t.dtypes['c'] == 'float64'
assert t.rows == 2
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)
# harder CSV - promote to object during inference period,
# duplicate column name
t = Tafra.read_csv('test/ex3.csv')
assert t.dtypes['a'] == 'int32'
assert t.dtypes['b'] == 'object'
assert t.dtypes['b (2)'] == 'float64'
assert t.rows == 6
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)
# as above, but with a promotion required after inference period
# (heuristic fails)
t = Tafra.read_csv('test/ex4.csv')
assert t.dtypes['a'] == 'int32'
assert t.dtypes['b'] == 'object'
assert t.dtypes['b (2)'] == 'float64'
assert t.rows == 6
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)
# bad CSV - missing column on row #4
with pytest.raises(ValueError) as e:
t = Tafra.read_csv('test/ex5.csv')
# bad CSV - missing column on row #4 - after guess rows
with pytest.raises(ValueError) as e:
t = Tafra.read_csv('test/ex5.csv', guess_rows=2)
# missing column - but numpy will automatically convert missing (None) to nan
t = Tafra.read_csv('test/ex6.csv')
assert t.dtypes['dp'] == 'float64'
assert t.dtypes['dp_prime'] == 'float64'
assert t.dtypes['dp_prime_te'] == 'float64'
assert t.dtypes['t'] == 'float64'
assert t.dtypes['te'] == 'float64'
check_tafra(t)
# missing column - do not automatically cast
t = Tafra.read_csv('test/ex6.csv', missing=None)
assert t.dtypes['dp'] == 'float64'
assert t.dtypes['dp_prime'] == 'object'
assert t.dtypes['dp_prime_te'] == 'object'
assert t.dtypes['t'] == 'float64'
assert t.dtypes['te'] == 'float64'
check_tafra(t)
t.update_dtypes_inplace({'dp_prime': float, 'dp_prime_te': 'float64'})
assert t.dtypes['dp_prime'] == 'float64'
assert t.dtypes['dp_prime_te'] == 'float64'
check_tafra(t)
# force dtypes on missing columns
t = Tafra.read_csv('test/ex6.csv',
missing=None,
dtypes={
'dp_prime': np.float,
'dp_prime_te': np.float32
})
assert t.dtypes['dp'] == 'float64'
assert t.dtypes['dp_prime'] == 'float64'
assert t.dtypes['dp_prime_te'] == 'float32'
assert t.dtypes['t'] == 'float64'
assert t.dtypes['te'] == 'float64'
check_tafra(t)
# override a column type
t = Tafra.read_csv('test/ex4.csv', dtypes={'a': 'float32'})
assert t.dtypes['a'] == 'float32'
assert t.dtypes['b'] == 'object'
assert t.dtypes['b (2)'] == 'float64'
assert t.rows == 6
assert len(t.columns) == 3
check_tafra(t)
write_reread(t)