def _set_columns(self, columns):
names_seen = set()
for column in columns:
if not isinstance(column, tuple):
raise _user_exceptions.FlyteValueException(
column,
"When specifying a Schema type with a known set of columns. Each column must be "
"specified as a tuple in the form ('name', type).")
if len(column) != 2:
raise _user_exceptions.FlyteValueException(
column,
"When specifying a Schema type with a known set of columns. Each column must be "
"specified as a tuple in the form ('name', type).")
name, sdk_type = column
sdk_type = _helpers.python_std_to_sdk_type(sdk_type)
if not isinstance(name, (str, _six.text_type)):
additional_msg = "When specifying a Schema type with a known set of columns, the first element in" \
" each tuple must be text."
raise _user_exceptions.FlyteTypeException(
received_type=type(name),
received_value=name,
expected_type={str, _six.text_type},
additional_msg=additional_msg)
if not isinstance(sdk_type, _base_sdk_types.FlyteSdkType) or sdk_type.to_flyte_literal_type() not in \
get_supported_literal_types_to_pandas_types():
additional_msg = \
"When specifying a Schema type with a known set of columns, the second element of " \
"each tuple must be a supported type. Failed for column: {name}".format(
name=name)
raise _user_exceptions.FlyteTypeException(
expected_type=list(
get_supported_literal_types_to_pandas_types().keys()),
received_type=sdk_type,
additional_msg=additional_msg)
if name in names_seen:
raise ValueError(
"The column name {name} was specified multiple times when instantiating the "
"Schema.".format(name=name))
names_seen.add(name)
self._sdk_columns = _collections.OrderedDict(columns)
def infer_sdk_type_from_literal(literal):
"""
:param flytekit.models.literals.Literal literal:
:rtype: flytekit.common.types.base_sdk_types.FlyteSdkType
"""
for e in _TypeEngineLoader.iterate_engines_in_order():
out = e.infer_sdk_type_from_literal(literal)
if out is not None:
return out
raise _user_exceptions.FlyteValueException(literal, "Could not resolve to a type implementation for this value.")
def python_std_to_sdk_type(t):
"""
:param T t: User input. Should be of the form: Types.Integer, [Types.Integer], {Types.String: Types.Integer}, etc.
:rtype: flytekit.common.types.base_sdk_types.FlyteSdkType
"""
for e in _TypeEngineLoader.iterate_engines_in_order():
out = e.python_std_to_sdk_type(t)
if out is not None:
return out
raise _user_exceptions.FlyteValueException(t, "Could not resolve to an SDK type for this value.")
def from_string(cls, string_value):
"""
:param Text string_value:
:rtype: Blob
"""
if not string_value:
_user_exceptions.FlyteValueException(
string_value,
"Cannot create a Blob from the provided path value.")
return cls(_blob_impl.Blob.from_string(string_value, mode='rb'))
def from_string(cls, string_value):
"""
:param Text string_value:
:rtype: Schema
"""
if not string_value:
_user_exceptions.FlyteValueException(
string_value, "Cannot create a Schema from an empty path")
return cls(
_schema_impl.Schema.from_string(string_value,
schema_type=cls.schema_type))
def from_string(cls, string_value):
"""
:param Text string_value: Should be a JSON formatted string
:rtype: Generic
"""
try:
t = _json_format.Parse(string_value, _struct.Struct())
except Exception:
raise _user_exceptions.FlyteValueException(
string_value, "Could not be parsed from JSON.")
return cls(t)
def from_string(cls, string_value, schema_type=None):
"""
:param Text string_value:
:param SchemaType schema_type:
:rtype: Schema
"""
if not string_value:
_user_exceptions.FlyteValueException(
string_value, "Cannot create a Schema from an empty path")
return cls.create_at_known_location(string_value,
schema_type=schema_type)
def from_string(cls, string_value):
"""
:param Text string_value: b64 encoded string of bytes
:rtype: Protobuf
"""
try:
decoded = _base64.b64decode(string_value)
except TypeError:
raise _user_exceptions.FlyteValueException(string_value, "The string is not valid base64-encoded.")
pb_obj = cls.pb_type()
pb_obj.ParseFromString(decoded)
return cls(pb_obj)
def from_urn(cls, string: str) -> "WorkflowExecutionIdentifier":
"""
Parses a string in the correct format into an identifier
"""
segments = string.split(":")
if len(segments) != 4:
raise _user_exceptions.FlyteValueException(
string,
"The provided string was not in a parseable format. The string for an identifier must be in the format"
" ex:project:domain:name.",
)
resource_type, project, domain, name = segments
if resource_type != "ex":
raise _user_exceptions.FlyteValueException(
resource_type,
"The provided string could not be parsed. The first element of an execution identifier must be 'ex'.",
)
return cls(project, domain, name)
def _validate_phases(self, phases):
"""
:param list[int] phases:
"""
if len(phases) == 0:
raise _user_exceptions.FlyteAssertion("You must specify at least one phase for a notification.")
for phase in phases:
if phase not in self.VALID_PHASES:
raise _user_exceptions.FlyteValueException(
phase,
self.VALID_PHASES,
additional_message="Notifications can only be specified on terminal states."
)
def from_string(cls, string_value):
"""
:param Text string_value:
:rtype: MultiPartCSV
"""
if not string_value:
_user_exceptions.FlyteValueException(
string_value,
"Cannot create a MultiPartCSV from the provided path value.")
return cls(
_blob_impl.MultiPartBlob.from_string(string_value,
format='csv',
mode='r'))
def from_urn(cls, urn: str) -> "Identifier":
"""
Parses a string urn in the correct format into an identifier
"""
segments = urn.split(":")
if len(segments) != 5:
raise _user_exceptions.FlyteValueException(
urn,
"The provided string was not in a parseable format. The string for an identifier must be in the "
"format entity_type:project:domain:name:version.",
)
resource_type, project, domain, name, version = segments
if resource_type not in cls._STRING_TO_TYPE_MAP:
raise _user_exceptions.FlyteValueException(
resource_type,
"The provided string could not be parsed. The first element of an identifier must be one of: "
f"{list(cls._STRING_TO_TYPE_MAP.keys())}. ",
)
return cls(cls._STRING_TO_TYPE_MAP[resource_type], project, domain,
name, version)
def download(self, from_path, to_path):
"""
:param Text from_path:
:param Text to_path:
"""
rsp = _requests.get(from_path)
if rsp.status_code != type(self)._HTTP_OK:
raise _user_exceptions.FlyteValueException(
rsp.status_code,
"Request for data @ {} failed. Expected status code {}".format(from_path, type(self)._HTTP_OK),
)
with open(to_path, "wb") as writer:
writer.write(rsp.content)
def from_python_std(cls, string):
"""
Parses a string in the correct format into an identifier
:param Text string:
:rtype: Identifier
"""
segments = string.split(":")
if len(segments) != 5:
raise _user_exceptions.FlyteValueException(
"The provided string was not in a parseable format. The string for an identifier must be in the format"
" entity_type:project:domain:name:version. Received: {}".format(string)
)
resource_type, project, domain, name, version = segments
if resource_type not in cls._STRING_TO_TYPE_MAP:
raise _user_exceptions.FlyteValueException(
"The provided string could not be parsed. The first element of an identifier must be one of: {}. "
"Received: {}".format(list(cls._STRING_TO_TYPE_MAP.keys()), resource_type)
)
resource_type = cls._STRING_TO_TYPE_MAP[resource_type]
return cls(resource_type, project, domain, name, version)
def schema_instantiator(columns=None):
"""
:param list[(Text, flytekit.common.types.base_sdk_types.FlyteSdkType)] columns: [Optional] Description of the
columns in the underlying schema. Should be tuples with the first element being the name.
:rtype: SchemaInstantiator
"""
if columns is not None and len(columns) == 0:
raise _user_exceptions.FlyteValueException(
columns,
"When specifying a Schema type with a known set of columns, a non-empty list must be provided as "
"inputs")
class _Schema(_six.with_metaclass(SchemaInstantiator, Schema)):
_schema_type = _schema_impl.SchemaType(columns=columns)
return _Schema
def exists(self, path):
"""
:param Text path: the path of the file
:rtype bool: whether the file exists or not
"""
rsp = _requests.head(path)
allowed_codes = {
type(self)._HTTP_OK,
type(self)._HTTP_NOT_FOUND,
type(self)._HTTP_FORBIDDEN
}
if rsp.status_code not in allowed_codes:
raise _user_exceptions.FlyteValueException(
rsp.status_code,
"Data at {} could not be checked for existence. Expected one of: {}"
.format(path, allowed_codes))
return rsp.status_code == type(self)._HTTP_OK
def from_python_std(cls, t_value):
"""
:param T t_value: It is up to each individual object as to whether or not this value can be cast.
:rtype: FlyteSdkValue
:raises: flytekit.common.exceptions.user.FlyteTypeException
"""
if t_value is None:
return _base_sdk_types.Void()
elif type(t_value) != _datetime.datetime:
raise _user_exceptions.FlyteTypeException(type(t_value),
_datetime.datetime,
t_value)
elif t_value.tzinfo is None:
raise _user_exceptions.FlyteValueException(
t_value, "Datetime objects in Flyte must be timezone aware. "
"tzinfo was found to be None.")
return cls(t_value)
def from_python_std(cls, t_value):
"""
:param T t_value: It is up to each individual object as to whether or not this value can be cast.
:rtype: FlyteSdkValue
:raises: flytekit.common.exceptions.user.FlyteTypeException
"""
if t_value is None:
return _base_sdk_types.Void()
elif isinstance(t_value, _blob_impl.Blob):
if t_value.metadata.type.format != "csv":
raise _user_exceptions.FlyteValueException(
t_value, "Blob is in incorrect format. Expected CSV.")
blob = t_value
else:
blob = _blob_impl.Blob.from_python_std(t_value,
format="csv",
mode="w")
return cls(blob)
def to_literal_model(self):
"""
Converts current binding data into a Literal asserting that there are no promises in the bindings.
:rtype: Literal
"""
if self.promise:
raise _user_exceptions.FlyteValueException(
self.promise, "Cannot convert BindingData to a Literal because " "it has a promise.",
)
elif self.scalar:
return Literal(scalar=self.scalar)
elif self.collection:
return Literal(
collection=LiteralCollection(
literals=[binding.to_literal_model() for binding in self.collection.bindings]
)
)
elif self.map:
return Literal(map=LiteralMap(literals={k: binding.to_literal_model() for k, binding in self.map.bindings}))
def from_python_std(cls, t_value):
"""
:param T t_value: It is up to each individual object as to whether or not this value can be cast.
:rtype: FlyteSdkValue
:raises: flytekit.common.exceptions.user.FlyteTypeException
"""
if t_value is None:
return _base_sdk_types.Void()
elif not isinstance(t_value, dict):
raise _user_exceptions.FlyteTypeException(type(t_value), dict,
t_value)
try:
t = _json.dumps(t_value)
except Exception:
raise _user_exceptions.FlyteValueException(
t_value, "Is not JSON serializable.")
return cls(_json_format.Parse(t, _struct.Struct()))
def _load_engines(cls):
config = _sdk_config.TYPE_ENGINES.get()
if cls._LOADED_ENGINES is None or config != cls._LAST_LOADED:
cls._LAST_LOADED = config
cls._LOADED_ENGINES = []
for fqdn in config:
split = fqdn.split(".")
module_path, attr = ".".join(split[:-1]), split[-1]
module = _exception_scopes.user_entry_point(
_importlib.import_module)(module_path)
if not hasattr(module, attr):
raise _user_exceptions.FlyteValueException(
module,
"Failed to load the type engine because the attribute named '{}' could not be found"
"in the module '{}'.".format(attr, module_path))
engine_impl = getattr(module, attr)()
cls._LOADED_ENGINES.append(engine_impl)
from flytekit.type_engines.default.flyte import FlyteDefaultTypeEngine as _DefaultEngine
cls._LOADED_ENGINES.append(_DefaultEngine())
def _load_type_from_tag(tag: str) -> Type:
"""
Loads python type from tag
"""
if "." not in tag:
raise _user_exceptions.FlyteValueException(
tag, "Protobuf tag must include at least one '.' to delineate package and object name.",
)
module, name = tag.rsplit(".", 1)
try:
pb_module = _importer.import_module(module)
except ImportError:
raise _user_exceptions.FlyteAssertion(
"Could not resolve the protobuf definition @ {}. Is the protobuf library installed?".format(module)
)
if not hasattr(pb_module, name):
raise _user_exceptions.FlyteAssertion("Could not find the protobuf named: {} @ {}.".format(name, module))
return getattr(pb_module, name)
def _proto_sdk_type_from_tag(tag):
"""
:param Text tag:
:rtype: _proto.Protobuf
"""
if "." not in tag:
raise _user_exceptions.FlyteValueException(
tag, "Protobuf tag must include at least one '.' to delineate package and object name.",
)
module, name = tag.rsplit(".", 1)
try:
pb_module = _importer.import_module(module)
except ImportError:
raise _user_exceptions.FlyteAssertion(
"Could not resolve the protobuf definition @ {}. Is the protobuf library installed?".format(module)
)
if not hasattr(pb_module, name):
raise _user_exceptions.FlyteAssertion("Could not find the protobuf named: {} @ {}.".format(name, module))
return _proto.create_protobuf(getattr(pb_module, name))
def _discover_workflow_components(workflow_class):
"""
This task iterates over the attributes of a user-defined class in order to return a list of inputs, outputs and
nodes.
:param class workflow_class: User-defined class with task instances as attributes.
:rtype: (list[flytekit.common.promise.Input], list[Output], list[flytekit.common.nodes.SdkNode])
"""
inputs = []
outputs = []
nodes = []
to_visit_objs = _queue.Queue()
top_level_attributes = set()
for attribute_name in dir(workflow_class):
to_visit_objs.put(
(attribute_name, getattr(workflow_class, attribute_name)))
top_level_attributes.add(attribute_name)
# For all task instances defined within the workflow, bind them to this specific workflow and hook-up to the
# engine (when available)
visited_obj_ids = set()
while not to_visit_objs.empty():
attribute_name, current_obj = to_visit_objs.get()
current_obj_id = id(current_obj)
if current_obj_id in visited_obj_ids:
continue
visited_obj_ids.add(current_obj_id)
if isinstance(current_obj, _nodes.SdkNode):
# TODO: If an attribute name is on the form node_name[index], the resulting
# node name might not be correct.
nodes.append(current_obj.assign_id_and_return(attribute_name))
elif isinstance(current_obj, _promise.Input):
if attribute_name is None or attribute_name not in top_level_attributes:
raise _user_exceptions.FlyteValueException(
attribute_name,
"Detected workflow input specified outside of top level.",
)
inputs.append(
current_obj.rename_and_return_reference(attribute_name))
elif isinstance(current_obj, Output):
if attribute_name is None or attribute_name not in top_level_attributes:
raise _user_exceptions.FlyteValueException(
attribute_name,
"Detected workflow output specified outside of top level.",
)
outputs.append(
current_obj.rename_and_return_reference(attribute_name))
elif isinstance(current_obj, list) or isinstance(
current_obj, set) or isinstance(current_obj, tuple):
for idx, value in enumerate(current_obj):
to_visit_objs.put(
(_assign_indexed_attribute_name(attribute_name,
idx), value))
elif isinstance(current_obj, dict):
# Visit dictionary keys.
for key in current_obj.keys():
to_visit_objs.put(
(_assign_indexed_attribute_name(attribute_name, key), key))
# Visit dictionary values.
for key, value in _six.iteritems(current_obj):
to_visit_objs.put(
(_assign_indexed_attribute_name(attribute_name,
key), value))
return inputs, outputs, nodes
def _content_type_to_blob_format(content_type: _training_job_models) -> str:
if content_type == _training_job_models.InputContentType.TEXT_CSV:
return "csv"
else:
raise _user_exceptions.FlyteValueException(
"Unsupported InputContentType: {}".format(content_type))
