import inspect
from typing import (
Any,
NamedTuple,
Optional,
Type,
TypeVar,
Union,
)
import dagster._check as check
from dagster._annotations import PublicAttr, deprecated_param
from dagster._core.definitions.metadata import (
ArbitraryMetadataMapping,
RawMetadataMapping,
normalize_metadata,
)
from dagster._core.errors import DagsterError, DagsterInvalidDefinitionError
from dagster._core.types.dagster_type import (
DagsterType,
is_dynamic_output_annotation,
resolve_dagster_type,
)
from .inference import InferredOutputProps
from .input import NoValueSentinel
from .utils import DEFAULT_IO_MANAGER_KEY, DEFAULT_OUTPUT, check_valid_name
TOutputDefinition = TypeVar("TOutputDefinition", bound="OutputDefinition")
TOut = TypeVar("TOut", bound="Out")
class OutputDefinition:
"""Defines an output from an op's compute function.
Ops can have multiple outputs, in which case outputs cannot be anonymous.
Many ops have only one output, in which case the user can provide a single output definition
that will be given the default name, "result".
Output definitions may be typed using the Dagster type system.
Args:
dagster_type (Optional[Union[Type, DagsterType]]]): The type of this output.
Users should provide the Python type of the objects that they expect the op to yield
for this output, or a :py:class:`DagsterType` that defines a runtime check that they
want to be run on this output. Defaults to :py:class:`Any`.
name (Optional[str]): Name of the output. (default: "result")
description (Optional[str]): Human-readable description of the output.
is_required (Optional[bool]): Whether the presence of this field is required. (default: True)
io_manager_key (Optional[str]): The resource key of the IOManager used for storing this
output and loading it in downstream steps (default: "io_manager").
metadata (Optional[Dict[str, Any]]): A dict of the metadata for the output.
For example, users can provide a file path if the data object will be stored in a
filesystem, or provide information of a database table when it is going to load the data
into the table.
code_version (Optional[str]): (Experimental) Version of the code that generates this output. In
general, versions should be set only for code that deterministically produces the same
output when given the same inputs.
"""
def __init__(
self,
dagster_type=None,
name: Optional[str] = None,
description: Optional[str] = None,
is_required: bool = True,
io_manager_key: Optional[str] = None,
metadata: Optional[ArbitraryMetadataMapping] = None,
code_version: Optional[str] = None,
# make sure new parameters are updated in combine_with_inferred below
):
self._name = check_valid_name(check.opt_str_param(name, "name", DEFAULT_OUTPUT))
self._type_not_set = dagster_type is None
self._dagster_type = resolve_dagster_type(dagster_type)
self._description = check.opt_str_param(description, "description")
self._is_required = check.bool_param(is_required, "is_required")
self._io_manager_key = check.opt_str_param(
io_manager_key,
"io_manager_key",
default=DEFAULT_IO_MANAGER_KEY,
)
self._code_version = check.opt_str_param(code_version, "code_version")
self._raw_metadata = check.opt_mapping_param(metadata, "metadata", key_type=str)
self._metadata = normalize_metadata(self._raw_metadata, allow_invalid=True)
@property
def name(self) -> str:
return self._name
@property
def dagster_type(self) -> DagsterType:
return self._dagster_type
@property
def description(self) -> Optional[str]:
return self._description
@property
def is_required(self) -> bool:
return self._is_required
@property
def io_manager_key(self) -> str:
return self._io_manager_key
@property
def code_version(self) -> Optional[str]:
return self._code_version
@property
def optional(self) -> bool:
return not self.is_required
@property
def metadata(self) -> ArbitraryMetadataMapping:
return self._raw_metadata
@property
def is_dynamic(self) -> bool:
return False
def mapping_from(
self, node_name: str, output_name: Optional[str] = None, from_dynamic_mapping: bool = False
) -> "OutputMapping":
"""Create an output mapping from an output of a child node.
In a GraphDefinition, you can use this helper function to construct
an :py:class:`OutputMapping` from the output of a child node.
Args:
node_name (str): The name of the child node from which to map this output.
output_name (str): The name of the child node's output from which to map this output.
Examples:
.. code-block:: python
output_mapping = OutputDefinition(Int).mapping_from('child_node')
"""
return OutputMapping(
graph_output_name=self.name,
mapped_node_name=node_name,
mapped_node_output_name=output_name or DEFAULT_OUTPUT,
graph_output_description=self.description,
dagster_type=self.dagster_type,
from_dynamic_mapping=from_dynamic_mapping or self.is_dynamic,
)
@staticmethod
def create_from_inferred(
inferred: Optional[InferredOutputProps], code_version: Optional[str] = None
) -> "OutputDefinition":
if not inferred:
return OutputDefinition(code_version=code_version)
if is_dynamic_output_annotation(inferred.annotation):
return DynamicOutputDefinition(
dagster_type=_checked_inferred_type(inferred.annotation),
description=inferred.description,
code_version=code_version,
)
else:
return OutputDefinition(
dagster_type=_checked_inferred_type(inferred.annotation),
description=inferred.description,
code_version=code_version,
)
def combine_with_inferred(
self: TOutputDefinition, inferred: InferredOutputProps
) -> TOutputDefinition:
dagster_type = self.dagster_type
if self._type_not_set:
dagster_type = _checked_inferred_type(inferred.annotation)
if self.description is None:
description = inferred.description
else:
description = self.description
return self.__class__(
name=self.name,
dagster_type=dagster_type,
description=description,
is_required=self.is_required,
io_manager_key=self.io_manager_key,
metadata=self._metadata,
)
def _checked_inferred_type(inferred: Any) -> DagsterType:
try:
if inferred == inspect.Parameter.empty:
return resolve_dagster_type(None)
elif inferred is None:
# When inferred.annotation is None, it means someone explicitly put "None" as the
# annotation, so want to map it to a DagsterType that checks for the None type
return resolve_dagster_type(type(None))
else:
return resolve_dagster_type(inferred)
except DagsterError as e:
raise DagsterInvalidDefinitionError(
f"Problem using type '{inferred}' from return type annotation, correct the issue "
"or explicitly set the dagster_type via Out()."
) from e
class DynamicOutputDefinition(OutputDefinition):
"""Variant of :py:class:`OutputDefinition <dagster.OutputDefinition>` for an
output that will dynamically alter the graph at runtime.
When using in a composition function such as :py:func:`@job <dagster.job>`,
dynamic outputs must be used with either:
* ``map`` - clone downstream nodes for each separate :py:class:`DynamicOutput`
* ``collect`` - gather across all :py:class:`DynamicOutput` in to a list
Uses the same constructor as :py:class:`OutputDefinition <dagster.OutputDefinition>`
.. code-block:: python
@op(
config_schema={
"path": Field(str, default_value=file_relative_path(__file__, "sample"))
},
output_defs=[DynamicOutputDefinition(str)],
)
def files_in_directory(context):
path = context.op_config["path"]
dirname, _, filenames = next(os.walk(path))
for file in filenames:
yield DynamicOutput(os.path.join(dirname, file), mapping_key=_clean(file))
@job
def process_directory():
files = files_in_directory()
# use map to invoke an op on each dynamic output
file_results = files.map(process_file)
# use collect to gather the results in to a list
summarize_directory(file_results.collect())
"""
@property
def is_dynamic(self) -> bool:
return True
class OutputPointer(NamedTuple("_OutputPointer", [("node_name", str), ("output_name", str)])):
def __new__(cls, node_name: str, output_name: Optional[str] = None):
return super(OutputPointer, cls).__new__(
cls,
check.str_param(node_name, "node_name"),
check.opt_str_param(output_name, "output_name", DEFAULT_OUTPUT),
)
[docs]@deprecated_param(
param="dagster_type",
breaking_version="2.0",
additional_warn_text="Any defined `dagster_type` should come from the underlying op `Output`.",
# Disabling warning here since we're passing this internally and I'm not sure whether it is
# actually used or discarded.
emit_runtime_warning=False,
)
class OutputMapping(NamedTuple):
"""Defines an output mapping for a graph.
Args:
graph_output_name (str): Name of the output in the graph being mapped to.
mapped_node_name (str): Named of the node (op/graph) that the output is being mapped from.
mapped_node_output_name (str): Name of the output in the node (op/graph) that is being mapped from.
graph_output_description (Optional[str]): A description of the output in the graph being mapped from.
from_dynamic_mapping (bool): Set to true if the node being mapped to is a mapped dynamic node.
dagster_type (Optional[DagsterType]): The dagster type of the graph's output being mapped to.
Examples:
.. code-block:: python
from dagster import OutputMapping, GraphDefinition, op, graph, GraphOut
@op
def emit_five(x):
return 5
# The following two graph definitions are equivalent
GraphDefinition(
name="the_graph",
node_defs=[emit_five],
output_mappings=[
OutputMapping(
graph_output_name="result", # Default output name
mapped_node_name="emit_five",
mapped_node_output_name="result"
)
]
)
@graph(out=GraphOut())
def the_graph:
return emit_five()
"""
graph_output_name: str
mapped_node_name: str
mapped_node_output_name: str
graph_output_description: Optional[str] = None
dagster_type: Optional[DagsterType] = None
from_dynamic_mapping: bool = False
@property
def maps_from(self) -> OutputPointer:
return OutputPointer(self.mapped_node_name, self.mapped_node_output_name)
def get_definition(self, is_dynamic: bool) -> "OutputDefinition":
check.invariant(not is_dynamic or self.from_dynamic_mapping)
is_dynamic = is_dynamic or self.from_dynamic_mapping
klass = DynamicOutputDefinition if is_dynamic else OutputDefinition
return klass(
name=self.graph_output_name,
description=self.graph_output_description,
dagster_type=self.dagster_type,
)
[docs]class Out(
NamedTuple(
"_Out",
[
("dagster_type", PublicAttr[Union[DagsterType, Type[NoValueSentinel]]]),
("description", PublicAttr[Optional[str]]),
("is_required", PublicAttr[bool]),
("io_manager_key", PublicAttr[str]),
("metadata", PublicAttr[Optional[RawMetadataMapping]]),
("code_version", PublicAttr[Optional[str]]),
],
)
):
"""Defines an output from an op's compute function.
Ops can have multiple outputs, in which case outputs cannot be anonymous.
Many ops have only one output, in which case the user can provide a single output definition
that will be given the default name, "result".
Outs may be typed using the Dagster type system.
Args:
dagster_type (Optional[Union[Type, DagsterType]]]):
The type of this output. Should only be set if the correct type can not
be inferred directly from the type signature of the decorated function.
description (Optional[str]): Human-readable description of the output.
is_required (bool): Whether the presence of this field is required. (default: True)
io_manager_key (Optional[str]): The resource key of the output manager used for this output.
(default: "io_manager").
metadata (Optional[Dict[str, Any]]): A dict of the metadata for the output.
For example, users can provide a file path if the data object will be stored in a
filesystem, or provide information of a database table when it is going to load the data
into the table.
code_version (Optional[str]): (Experimental) Version of the code that generates this output. In
general, versions should be set only for code that deterministically produces the same
output when given the same inputs.
"""
def __new__(
cls,
dagster_type: Optional[Union[Type, DagsterType]] = NoValueSentinel,
description: Optional[str] = None,
is_required: bool = True,
io_manager_key: Optional[str] = None,
metadata: Optional[ArbitraryMetadataMapping] = None,
code_version: Optional[str] = None,
# make sure new parameters are updated in combine_with_inferred below
):
return super(Out, cls).__new__(
cls,
dagster_type=(
NoValueSentinel
if dagster_type is NoValueSentinel
else resolve_dagster_type(dagster_type)
),
description=description,
is_required=check.bool_param(is_required, "is_required"),
io_manager_key=check.opt_str_param(
io_manager_key, "io_manager_key", default=DEFAULT_IO_MANAGER_KEY
),
metadata=metadata,
code_version=code_version,
)
@classmethod
def from_definition(cls, output_def: "OutputDefinition"):
klass = Out if not output_def.is_dynamic else DynamicOut
return klass(
dagster_type=output_def.dagster_type,
description=output_def.description,
is_required=output_def.is_required,
io_manager_key=output_def.io_manager_key,
metadata=output_def.metadata,
code_version=output_def.code_version,
)
def to_definition(
self,
annotation_type: type,
name: Optional[str],
description: Optional[str],
code_version: Optional[str],
) -> "OutputDefinition":
dagster_type = (
self.dagster_type
if self.dagster_type is not NoValueSentinel
else _checked_inferred_type(annotation_type)
)
klass = OutputDefinition if not self.is_dynamic else DynamicOutputDefinition
return klass(
dagster_type=dagster_type,
name=name,
description=self.description or description,
is_required=self.is_required,
io_manager_key=self.io_manager_key,
metadata=self.metadata,
code_version=self.code_version or code_version,
)
@property
def is_dynamic(self) -> bool:
return False
[docs]class DynamicOut(Out):
"""Variant of :py:class:`Out <dagster.Out>` for an output that will dynamically alter the graph at
runtime.
When using in a composition function such as :py:func:`@graph <dagster.graph>`,
dynamic outputs must be used with either
* ``map`` - clone downstream ops for each separate :py:class:`DynamicOut`
* ``collect`` - gather across all :py:class:`DynamicOut` in to a list
Uses the same constructor as :py:class:`Out <dagster.Out>`
.. code-block:: python
@op(
config_schema={
"path": Field(str, default_value=file_relative_path(__file__, "sample"))
},
out=DynamicOut(str),
)
def files_in_directory(context):
path = context.op_config["path"]
dirname, _, filenames = next(os.walk(path))
for file in filenames:
yield DynamicOutput(os.path.join(dirname, file), mapping_key=_clean(file))
@job
def process_directory():
files = files_in_directory()
# use map to invoke an op on each dynamic output
file_results = files.map(process_file)
# use collect to gather the results in to a list
summarize_directory(file_results.collect())
"""
def to_definition(
self,
annotation_type: type,
name: Optional[str],
description: Optional[str],
code_version: Optional[str],
) -> "OutputDefinition":
dagster_type = (
self.dagster_type
if self.dagster_type is not NoValueSentinel
else _checked_inferred_type(annotation_type)
)
return DynamicOutputDefinition(
dagster_type=dagster_type,
name=name,
description=self.description or description,
is_required=self.is_required,
io_manager_key=self.io_manager_key,
metadata=self.metadata,
code_version=self.code_version or code_version,
)
@property
def is_dynamic(self) -> bool:
return True
[docs]class GraphOut(NamedTuple("_GraphOut", [("description", PublicAttr[Optional[str]])])):
"""Represents information about the outputs that a graph maps.
Args:
description (Optional[str]): Human-readable description of the output.
"""
def __new__(cls, description: Optional[str] = None):
return super(GraphOut, cls).__new__(cls, description=description)
def to_definition(self, name: Optional[str]) -> "OutputDefinition":
return OutputDefinition(name=name, description=self.description)