def __init__(
self, source: Union[str, Source[Iterable[str]],
SimulatedEnvironment]) -> None:
if isinstance(source, str):
self._source = UrlSource(source)
elif isinstance(source, SimulatedEnvironment):
self._source = self._make_serialized_source(source)
else:
self._source = source
self._decoder = JsonDecode()
class EnvironmentDefinitionFileV1(Source[Sequence[Environment]]):
@overload
def __init__(self, filesource: Source[Iterable[str]]) -> None:
...
@overload
def __init__(self, filename: str) -> None:
...
def __init__(self, arg) -> None:
self._source = UrlSource(arg) if isinstance(arg, str) else arg
def read(self) -> Sequence[Environment]:
definitions: dict = JsonDecode().filter('\n'.join(self._source.read()))
variables = {
k: CobaRegistry.construct(v)
for k, v in definitions.get("variables", {}).items()
}
def _construct(item: Any) -> Sequence[Any]:
result = None
if isinstance(item, str) and item in variables:
result = variables[item]
if isinstance(item, str) and item not in variables:
result = CobaRegistry.construct(item)
if isinstance(item, dict):
result = CobaRegistry.construct(item)
if isinstance(item, list):
pieces = list(map(_construct, item))
if hasattr(pieces[0][0], 'read'):
result = [
Pipes.join(s, *f) for s in pieces[0]
for f in product(*pieces[1:])
]
else:
result = sum(pieces, [])
if result is None:
raise CobaException(
f"We were unable to construct {item} in the given environment definition file."
)
return result if isinstance(
result, collections.abc.Sequence) else [result]
if not isinstance(definitions['environments'], list):
definitions['environments'] = [definitions['environments']]
return [
environment for recipe in definitions['environments']
for environment in _construct(recipe)
]
def __init__(self, source: Union[str, Source[Iterable[str]]]) -> None:
"""Instantiate a LibsvmSource.
Args:
source: The data source. Accepts either a string representing the source location or another Source.
"""
source = UrlSource(source) if isinstance(source, str) else source
reader = LibsvmReader()
self._source = Pipes.join(source, reader)
class SerializedSimulation(SimulatedEnvironment):
def _make_serialized_source(
self, sim: SimulatedEnvironment) -> Source[Iterable[str]]:
def serialized_generator() -> Iterable[str]:
json_encoder = JsonEncode()
yield json_encoder.filter(sim.params)
for interaction in sim.read():
context = json_encoder.filter(interaction.context)
actions = json_encoder.filter(interaction.actions)
kwargs = json_encoder.filter(interaction.kwargs)
yield f"[{context},{actions},{kwargs}]"
return LambdaSource(serialized_generator)
def __init__(
self, source: Union[str, Source[Iterable[str]],
SimulatedEnvironment]) -> None:
if isinstance(source, str):
self._source = UrlSource(source)
elif isinstance(source, SimulatedEnvironment):
self._source = self._make_serialized_source(source)
else:
self._source = source
self._decoder = JsonDecode()
@property
def params(self) -> Dict[str, Any]:
return self._decoder.filter(next(iter(self._source.read())))
def read(self) -> Iterable[SimulatedInteraction]:
for interaction_json in islice(self._source.read(), 1, None):
deocded_interaction = self._decoder.filter(interaction_json)
yield SimulatedInteraction(deocded_interaction[0],
deocded_interaction[1],
**deocded_interaction[2])
def write(self, sink: Sink[str]):
for line in self._source.read():
sink.write(line)
def __init__(self,
source: Union[str, Source[Iterable[str]]],
has_header: bool = False,
**dialect) -> None:
"""Instantiate a CsvSource.
Args:
source: The data source. Accepts either a string representing the source location or another Source.
has_header: Indicates if the CSV files has a header row.
"""
source = UrlSource(source) if isinstance(source, str) else source
reader = CsvReader(has_header, **dialect)
self._source = Pipes.join(source, reader)
def __init__(self,
source: Union[str, Source[Iterable[str]]],
cat_as_str: bool = False,
skip_encoding: bool = False,
lazy_encoding: bool = True,
header_indexing: bool = True) -> None:
"""Instantiate an ArffSource.
Args:
source: The data source. Accepts either a string representing the source location or another Source.
cat_as_str: Indicates that categorical features should be encoded as a string rather than one hot encoded.
skip_encoding: Indicates that features should not be encoded (this means all features will be strings).
lazy_encoding: Indicates that features should be encoded lazily (this can save time if rows will be dropped).
header_indexing: Indicates that header data should be preserved so rows can be indexed by header name.
"""
source = UrlSource(source) if isinstance(source, str) else source
reader = ArffReader(cat_as_str, skip_encoding, lazy_encoding,
header_indexing)
self._source = Pipes.join(source, reader)
def test_unknown_scheme(self):
with self.assertRaises(CobaException):
UrlSource("irc://fail")
def test_no_scheme(self):
url = "c:/users"
self.assertIsInstance(UrlSource(url)._source, DiskSource)
self.assertEqual(url, UrlSource(url)._source._filename)
def test_file_scheme(self):
url = "file://c:/users"
self.assertIsInstance(UrlSource(url)._source, DiskSource)
self.assertEqual(url[7:], UrlSource(url)._source._filename)
def test_https_scheme(self):
url = "https://www.google.com"
self.assertIsInstance(UrlSource(url)._source, HttpSource)
self.assertEqual(url, UrlSource(url)._source._url)
def __init__(self, arg) -> None:
self._source = UrlSource(arg) if isinstance(arg, str) else arg