def test_formats_serdes():
objects = [
{
'blah': 'foo'
},
b'blather',
'blip',
]
metadata = [{} for o in objects]
for obj, meta in zip(objects, metadata):
data, format_meta = FormatRegistry.serialize(obj, meta)
meta.update(format_meta)
assert FormatRegistry.deserialize(data, meta) == obj
meta = {}
df1 = pd.DataFrame([[1, 2], [3, 4]])
data, format_meta = FormatRegistry.serialize(df1, meta)
meta.update(format_meta)
df2 = FormatRegistry.deserialize(data, meta)
# we can't really get around this nicely -- if header is used, and header names are numeric,
# once loaded from CSV, header names are now strings. This causes a bad comparison, so we
# cast to int again.
df2.columns = df2.columns.astype(int, copy=False)
assert df1.equals(df2)
def test_formats_csv_roundtrip():
test_data = b'9,2,5\n7,2,6\n1,0,1\n'
# roundtrip defaults.
meta = {'format': {'name': 'csv'}}
df1 = FormatRegistry.deserialize(test_data, meta)
bin, format_meta = FormatRegistry.serialize(df1, meta)
meta.update(format_meta)
df2 = FormatRegistry.deserialize(bin, meta)
assert test_data == bin
assert df1.equals(df2)
# interpret first row as header
meta = {'format': {'name': 'csv', 'opts': {'use_header': True}}}
df1 = FormatRegistry.deserialize(test_data, meta)
bin, format_meta = FormatRegistry.serialize(df1, meta)
meta.update(format_meta)
df2 = FormatRegistry.deserialize(bin, meta)
assert test_data == bin
assert df1.equals(df2)
# interpret first column as index
meta = {'format': {'name': 'csv', 'opts': {'use_index': True}}}
df1 = FormatRegistry.deserialize(test_data, meta)
bin, format_meta = FormatRegistry.serialize(df1, meta)
meta.update(format_meta)
df2 = FormatRegistry.deserialize(bin, meta)
assert test_data == bin
assert df1.equals(df2)
# interpret first row as header, and first column as index
meta = {
'format': {
'name': 'csv',
'opts': {
'use_index': True,
'use_header': True
}
}
}
df1 = FormatRegistry.deserialize(test_data, meta)
bin, format_meta = FormatRegistry.serialize(df1, meta)
meta.update(format_meta)
df2 = FormatRegistry.deserialize(bin, meta)
assert test_data == bin
assert df1.equals(df2)
def test_formats_csv_read():
csv_file = pathlib.Path(__file__).parent / 'data' / 'csv.csv'
meta = {'format': {'name': 'csv'}}
expected_bytes = b'a,b,c,d\n1,2,3,4\n5,6,7,8\n'
expected_df = FormatRegistry.deserialize(expected_bytes, meta)
df = FormatRegistry.deserialize(csv_file.read_bytes(), meta)
assert df.equals(expected_df)
assert expected_bytes == FormatRegistry.serialize(df, meta)[0]