def test_reserve_dtype(self):
t = TypedDfBuilder("a").reserve("x", dtype=np.float32).build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="0.5"))]))
assert df.column_names() == ["x"]
assert df.to_numpy().tolist() == [[0.5]]
with pytest.raises(ValueError):
t.convert(pd.DataFrame([pd.Series(dict(x="kitten"))]))
def test_to_instances(self):
t = TypedDfBuilder("T").require("animal", dtype=str).reserve("age", dtype=int).build()
df: t = t.of(
[
pd.Series(dict(animal="goldfish", age=2)),
pd.Series(dict(animal="goldfish", age=1)),
pd.Series(dict(animal="gazelle", age=8)),
pd.Series(dict(animal="pineapple", age=114)),
pd.Series(dict(animal="anteater", age=11)),
]
)
dc = t.create_dataclass()
instances = df.to_dataclass_instances()
assert instances == [
dc("goldfish", 2),
dc("goldfish", 1),
dc("gazelle", 8),
dc("pineapple", 114),
dc("anteater", 11),
]
assert list(sorted(instances)) == [
dc("anteater", 11),
dc("gazelle", 8),
dc("goldfish", 1),
dc("goldfish", 2),
dc("pineapple", 114),
]
def test_no_overwrite(self):
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat", y="dog"))]))
with tmpfile(".csv") as path:
df.write_file(path, overwrite=False)
with pytest.raises(FileExistsError):
df.write_file(path, overwrite=False)
def test_bad_type(self):
with pytest.raises(TypeError):
# noinspection PyTypeChecker
TypedDfBuilder(None).build()
with pytest.raises(TypeError):
# noinspection PyTypeChecker
TypedDfBuilder(5).build()
def test_mkdir(self):
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat", y="dog"))]))
with tmpdir() as path:
df.write_file(path / "a.csv", mkdirs=True)
with tmpdir() as path:
with pytest.raises(FileNotFoundError):
df.write_file(path / "b.csv")
def test_bad_reserve(self):
for index in [True, False]:
with pytest.raises(ClashError):
TypedDfBuilder("a").reserve("level_0", index=index)
with pytest.raises(ClashError):
TypedDfBuilder("a").reserve("abc", "level_0", index=index)
with pytest.raises(ClashError):
TypedDfBuilder("a").reserve("abc", "index", index=index)
def test_drop(self):
t = TypedDfBuilder("a").reserve("column").drop("trash").build()
typ: DfTyping = t.get_typing()
assert typ.columns_to_drop == {"trash"}
df = t.convert(pd.DataFrame([pd.Series(dict(x="x", zz="y"))]))
assert df.column_names() == ["x", "zz"]
df = t.convert(pd.DataFrame([pd.Series(dict(x="x", trash="y"))]))
assert df.column_names() == ["x"]
def test_require_and_reserve_col(self):
t = TypedDfBuilder("a").require("column").reserve("reserved").build()
typ: DfTyping = t.get_typing()
assert typ.required_columns == ["column"]
assert typ.reserved_columns == ["reserved"]
assert typ.required_index_names == []
assert typ.reserved_index_names == []
assert typ.verifications == []
def test_pass_io_options(self):
t = TypedDfBuilder("a").reserve("x",
"y").add_write_kwargs(FileFormat.csv,
sep="&").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat", y="dog"))]))
with tmpfile(".csv") as path:
df.write_file(path)
lines = path.read_text(encoding="utf8").splitlines()
assert lines == ["x&y", "cat&dog"]
def test_read_instances_empty_fields(self):
@dataclass(frozen=True)
class Dc:
pass
t = TypedDfBuilder("T").reserve("animal", dtype=str).build()
df = t.from_dataclass_instances([Dc()])
assert len(df) == 1
assert "animal" not in df.columns
def test_create_any(self):
t = TypedDfBuilder("T").require("greeting").reserve("bonus").build()
dc = t.create_dataclass()
assert dc is not None
assert issubclass(dc, TypedDfDataclass)
assert len(dc.get_fields()) == 2
assert dc.get_fields()[0].name == "greeting"
assert dc.get_fields()[0].type == Any
assert dc.get_fields()[1].name == "bonus"
assert dc.get_fields()[1].type == Optional[Any]
def test_read_instances(self):
@dataclass(frozen=True)
class Dc:
animal: str
val: Optional[int]
t = TypedDfBuilder("T").require("animal", dtype=str).reserve("age", dtype=int).build()
df = t.from_dataclass_instances([Dc("cat", 1), Dc("kitten", 2)])
assert len(df) == 2
assert df.to_numpy().tolist() == [["cat", 1], ["kitten", 2]]
def test_attrs_hard(self):
meta = None
try:
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat",
y="kitten"))]))
df.attrs["matrix"] = np.zeros((2, 2))
with tmpfile(".csv") as path:
df.write_file(path, attrs=True)
meta = Path(str(path) + ".attrs.json")
assert meta.exists()
df = t.read_file(path, attrs=True)
assert df.attrs == {"matrix": [["0.0", "0.0"], ["0.0", "0.0"]]}
finally:
if meta is not None:
meta.unlink(missing_ok=True)
def test_attrs(self):
meta = None
try:
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat",
y="kitten"))]))
df.attrs["fruit"] = "apple"
with tmpfile(".csv") as path:
df.write_file(path, attrs=True)
meta = Path(str(path) + ".attrs.json")
assert meta.exists()
data = meta.read_text(encoding="utf-8").replace("\n",
"").replace(
" ", "")
assert data == '{"fruit": "apple"}'
df = t.read_file(path, attrs=True)
assert df.attrs == {"fruit": "apple"}
finally:
if meta is not None:
meta.unlink(missing_ok=True)
def test_condition(self):
t = TypedDfBuilder("a").verify(always_ok).build()
typ: DfTyping = t.get_typing()
assert typ.required_columns == []
assert typ.required_index_names == []
assert typ.verifications == [always_ok]
TypedDf(pd.DataFrame())
t = TypedDfBuilder("a").verify(always_fail).build()
with pytest.raises(VerificationFailedError):
t.convert(pd.DataFrame())
def test_require_and_reserve_index(self):
t = (TypedDfBuilder("a").require("column", index=True).reserve(
"reserved", index=True)).build()
typ: DfTyping = t.get_typing()
assert typ.required_columns == []
assert typ.reserved_columns == []
assert typ.required_index_names == ["column"]
assert typ.reserved_index_names == ["reserved"]
assert typ.known_index_names == ["column", "reserved"]
assert typ.known_column_names == []
assert typ.known_names == ["column", "reserved"]
assert typ.verifications == []
def test_already_added(self):
for cola in [True, False]:
for indexa in [True, False]:
for colb in [True, False]:
for indexb in [True, False]:
builder = TypedDfBuilder("a")
if cola:
builder = builder.require("a", index=indexa)
else:
cola = builder.reserve("a", index=indexa)
with pytest.raises(ClashError):
if colb:
builder.require("a", index=indexb)
else:
builder.reserve("a", index=indexb)
def typed(cls, name: str, doc: Optional[str] = None) -> TypedDfBuilder:
"""
Creates a new type with flexible requirements.
The class will enforce constraints and subclass :class:`typeddfs.typed_dfs.TypedDf`.
Args:
name: The name that will be used for the new class
doc: The docstring for the new class
Returns:
A builder instance (builder pattern) to be used with chained calls
Example:
``TypedDfs.typed("MyClass").require("name", index=True).build()``
"""
return TypedDfBuilder(name, doc)
def test_dir_hash(self):
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat", y="kitten"))]))
with tmpfile(".csv") as path:
hash_dir = Checksums().get_dirsum_of_file(path)
hash_dir.unlink(missing_ok=True)
df.write_file(path, dir_hash=True)
assert hash_dir.exists()
got = Checksums().load_dirsum_exact(hash_dir)
assert list(got.keys()) == [path]
hit = got[path]
assert len(hit) == 64
t.read_file(path, dir_hash=True)
t.read_file(path, hex_hash=hit)
def test_file_hash(self):
t = TypedDfBuilder("a").reserve("x", "y").build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="cat", y="dog"))]))
# unfortunately, the file that gets output is os-dependent
# \n vs \r\n is an issue, so we can't check the exact hash
with tmpfile(".csv") as path:
df.write_file(path, file_hash=True)
hash_file = Checksums().get_filesum_of_file(path)
assert hash_file.exists()
got = Checksums().load_filesum_of_file(path)
assert got.file_path == path
hit = got.hash_value
assert len(hit) == 64
t.read_file(path, file_hash=True)
t.read_file(path, hex_hash=hit)
def test_dtype_post_process(self):
# make sure these happen in the right order:
# 1. dtype conversions
# 2. post-processing
# 3. final conditions
def post(dd: BaseDf) -> BaseDf:
assert dd["x"].dtype == np.float32
dd2 = dd.copy()
dd2["x"] += 9
return dd2
def cond(dd: BaseDf):
return None if dd["x"].dtype == np.float32 else "failed"
t = (TypedDfBuilder("a").reserve(
"x", dtype=np.float32).post(post).verify(cond)).build()
df = t.convert(pd.DataFrame([pd.Series(dict(x="0.5"))]))
assert df.to_numpy().tolist() == [[9.5]]
def test_create_empty(self):
t = TypedDfBuilder("T").build()
dc = t.create_dataclass()
assert dc is not None
assert issubclass(dc, TypedDfDataclass)
assert dc.get_fields() == []
def test_strict(self):
# strict columns but not index
t = TypedDfBuilder("a").strict(index=False, cols=True).build()
typ: DfTyping = t.get_typing()
assert typ.more_indices_allowed
assert not typ.more_columns_allowed
t.convert(pd.DataFrame([pd.Series(dict(x="x"))]).set_index("x"))
with pytest.raises(UnexpectedColumnError):
t.convert(pd.DataFrame([pd.Series(dict(x="x"))]))
# strict index but not columns
t = TypedDfBuilder("a").strict(True, False).build()
typ: DfTyping = t.get_typing()
assert typ.more_columns_allowed
assert not typ.more_indices_allowed
t.convert(pd.DataFrame([pd.Series(dict(x="x"))]))
with pytest.raises(UnexpectedIndexNameError):
df = PrettyDf(
pd.DataFrame([pd.Series(dict(x="x"))]).set_index("x"))
assert df.index_names() == ["x"]
assert df.column_names() == []
t.convert(df)
# neither strict
t = TypedDfBuilder("a").strict(False, False).build()
t.convert(pd.DataFrame([pd.Series(dict(x="x"))]))
def test_read_write_insecure(self):
secure_type = TypedDfBuilder("a").secure().build()
bad_type = TypedDfBuilder("a").recommended_only().build()
with pytest.raises(UnsupportedOperationError):
# noinspection HttpUrlsUsage
secure_type.read_url("http://google.com") # nosec
secure = secure_type.new_df()
bad = bad_type.new_df()
for fmt in FileFormat:
for suffix in fmt.suffixes:
try:
with tmpfile(suffix) as path:
# should always complain about insecurity FIRST
if not fmt.is_secure:
with pytest.raises(FormatInsecureError):
secure_type.read_file(path)
with pytest.raises(FormatInsecureError):
secure.write_file(path)
path.unlink(missing_ok=True)
if not fmt.is_recommended:
with pytest.raises(FormatDiscouragedError):
bad_type.read_file(path)
with pytest.raises(FormatDiscouragedError):
bad.write_file(path)
except Exception:
logger.error(f"Failed on suffix {suffix}")
raise
def test_typed_subclass(self):
t1 = TypedDfBuilder("t1").build()
t2 = TypedDfBuilder("t2").subclass(t1).build()
assert issubclass(t2, t1)
assert not issubclass(t1, t2)
def test_drop_clash(self):
t = TypedDfBuilder("a").reserve("trash").drop("trash")
with pytest.raises(ClashError):
t.build()
def test_secure(self):
TypedDfBuilder("a").secure().hash(alg="sha256").build()
TypedDfBuilder("a").hash(alg="sha1").build()
with pytest.raises(DfTypeConstructionError):
TypedDfBuilder("a").secure().hash(alg="sha1").build()
def test_read_instances_empty(self):
t = TypedDfBuilder("T").require("animal", dtype=str).build()
df = t.from_dataclass_instances([])
assert len(df) == 0
def _get(name: str, t: Type[AbsDf] = PlainTypedDf) -> LazyDf:
url = f"https://raw.githubusercontent.com/mwaskom/seaborn-data/master/{name}.csv"
if t is None:
p, _, _ = FileFormat.split(url)
t = TypedDfBuilder(p.name).build()
return LazyDf.from_source(url, t)
def test_to_instances_empty(self):
t = TypedDfBuilder("T").reserve("animal", dtype=str).build()
df: t = t.of([])
instances = df.to_dataclass_instances()
assert instances == []