Named generic for dynamictableregions, etc.

This commit is contained in:
sneakers-the-rat 2024-07-26 19:54:35 -07:00
parent 3d8403e9e3
commit c98f25f973
Signed by untrusted user who does not match committer: jonny
GPG key ID: 6DCB96EF1E4D232D
7 changed files with 191 additions and 25 deletions

View file

@ -113,6 +113,28 @@ class Adapter(BaseModel):
Generate the corresponding linkML element for this adapter Generate the corresponding linkML element for this adapter
""" """
def get(self, name: str) -> Union[Group, Dataset]:
"""
Get the first item whose ``neurodata_type_def`` matches ``name``
Convenience wrapper around :meth:`.walk_field_values`
"""
return next(self.walk_field_values(self, 'neurodata_type_def', name))
def get_model_with_field(self, field: str) -> Generator[Union[Group, Dataset], None, None]:
"""
Yield models that have a non-None value in the given field.
Useful during development to find all the ways that a given
field is used.
Args:
field (str): Field to search for
"""
for model in self.walk_types(self, (Group, Dataset)):
if getattr(model, field, None) is not None:
yield model
def walk( def walk(
self, input: Union[BaseModel, dict, list] self, input: Union[BaseModel, dict, list]
) -> Generator[Union[BaseModel, Any, None], None, None]: ) -> Generator[Union[BaseModel, Any, None], None, None]:

View file

@ -248,7 +248,7 @@ class ClassAdapter(Adapter):
range="string", range="string",
) )
else: else:
name_slot = SlotDefinition(name="name", required=True, range="string", identifier=True) name_slot = SlotDefinition(name="name", required=True, range="string")
return name_slot return name_slot
def build_self_slot(self) -> SlotDefinition: def build_self_slot(self) -> SlotDefinition:

View file

@ -530,6 +530,45 @@ class MapArrayLikeAttributes(DatasetMap):
return res return res
class MapClassRange(DatasetMap):
"""
Datasets that are a simple named reference to another type without any
additional modification to that type.
"""
@classmethod
def check(c, cls: Dataset) -> bool:
"""
Check that we are a dataset with a ``neurodata_type_inc`` and a name but nothing else
"""
return (
cls.neurodata_type_inc
and not cls.neurodata_type_def
and not cls.attributes
and not cls.dims
and not cls.shape
and not cls.dtype
and cls.name
)
@classmethod
def apply(
c, cls: Dataset, res: Optional[BuildResult] = None, name: Optional[str] = None
) -> BuildResult:
"""
Replace the base class with a slot with an annotation that indicates
it should use the :class:`.Named` generic when generated to pydantic
"""
this_slot = SlotDefinition(
name=cls.name,
description=cls.doc,
range=f"{cls.neurodata_type_inc}",
annotations=[{'named': True}],
**QUANTITY_MAP[cls.quantity],
)
res = BuildResult(slots=[this_slot])
return res
# -------------------------------------------------- # --------------------------------------------------
# DynamicTable special cases # DynamicTable special cases
# -------------------------------------------------- # --------------------------------------------------

View file

@ -62,6 +62,7 @@ from pydantic import BaseModel
from nwb_linkml.maps import flat_to_nptyping from nwb_linkml.maps import flat_to_nptyping
from nwb_linkml.maps.naming import module_case, version_module_case from nwb_linkml.maps.naming import module_case, version_module_case
from nwb_linkml.includes import ModelTypeString, _get_name, NamedString, NamedImports
OPTIONAL_PATTERN = re.compile(r"Optional\[([\w\.]*)\]") OPTIONAL_PATTERN = re.compile(r"Optional\[([\w\.]*)\]")
@ -119,35 +120,16 @@ class NWBPydanticGenerator(PydanticGenerator):
- strip unwanted metadata - strip unwanted metadata
- generate range with any_of - generate range with any_of
""" """
for key in self.skip_meta: slot = AfterGenerateSlot.skip_meta(slot, self.skip_meta)
if key in slot.attribute.meta: slot = AfterGenerateSlot.make_array_anyofs(slot)
del slot.attribute.meta[key] slot = AfterGenerateSlot.make_named_class_range(slot)
# make array ranges in any_of
if "any_of" in slot.attribute.meta:
any_ofs = slot.attribute.meta["any_of"]
if all(["array" in expr for expr in any_ofs]):
ranges = []
is_optional = False
for expr in any_ofs:
# remove optional from inner type
pyrange = slot.attribute.range
is_optional = OPTIONAL_PATTERN.match(pyrange)
if is_optional:
pyrange = is_optional.groups()[0]
range_generator = NumpydanticArray(ArrayExpression(**expr["array"]), pyrange)
ranges.append(range_generator.make().range)
slot.attribute.range = "Union[" + ", ".join(ranges) + "]"
if is_optional:
slot.attribute.range = "Optional[" + slot.attribute.range + "]"
del slot.attribute.meta["any_of"]
return slot return slot
def before_render_template(self, template: PydanticModule, sv: SchemaView) -> PydanticModule: def before_render_template(self, template: PydanticModule, sv: SchemaView) -> PydanticModule:
if "source_file" in template.meta: if "source_file" in template.meta:
del template.meta["source_file"] del template.meta["source_file"]
return template
def compile_module( def compile_module(
self, module_path: Path = None, module_name: str = "test", **kwargs self, module_path: Path = None, module_name: str = "test", **kwargs
@ -169,6 +151,62 @@ class NWBPydanticGenerator(PydanticGenerator):
raise e raise e
class AfterGenerateSlot:
"""
Container class for slot-modification methods
"""
@staticmethod
def skip_meta(slot: SlotResult, skip_meta: tuple[str]) -> SlotResult:
for key in skip_meta:
if key in slot.attribute.meta:
del slot.attribute.meta[key]
return slot
@staticmethod
def make_array_anyofs(slot: SlotResult) -> SlotResult:
"""
Make a Union of array ranges if multiple array types specified in ``any_of``
"""
# make array ranges in any_of
if "any_of" in slot.attribute.meta:
any_ofs = slot.attribute.meta["any_of"]
if all(["array" in expr for expr in any_ofs]):
ranges = []
is_optional = False
for expr in any_ofs:
# remove optional from inner type
pyrange = slot.attribute.range
is_optional = OPTIONAL_PATTERN.match(pyrange)
if is_optional:
pyrange = is_optional.groups()[0]
range_generator = NumpydanticArray(ArrayExpression(**expr["array"]), pyrange)
ranges.append(range_generator.make().range)
slot.attribute.range = "Union[" + ", ".join(ranges) + "]"
if is_optional:
slot.attribute.range = "Optional[" + slot.attribute.range + "]"
del slot.attribute.meta["any_of"]
return slot
@staticmethod
def make_named_class_range(slot: SlotResult) -> SlotResult:
"""
When a slot has a ``named`` annotation, wrap it in :class:`.Named`
"""
if 'named' in slot.source.annotations and slot.source.annotations['named'].value:
slot.attribute.range = f"Named[{slot.attribute.range}]"
named_injects = [ModelTypeString, _get_name, NamedString]
if slot.injected_classes is None:
slot.injected_classes = named_injects
else:
slot.injected_classes.extend([ModelTypeString, _get_name, NamedString])
if slot.imports:
slot.imports += NamedImports
else:
slot.imports = NamedImports
return slot
def compile_python( def compile_python(
text_or_fn: str, package_path: Path = None, module_name: str = "test" text_or_fn: str, package_path: Path = None, module_name: str = "test"
) -> ModuleType: ) -> ModuleType:

View file

@ -0,0 +1,50 @@
"""
Classes and types that are injected in generated pydantic modules, but have no
corresponding representation in linkml (ie. that don't belong in :mod:`.lang_elements`
Used to customize behavior of pydantic classes either to match pynwb behavior or
reduce the verbosity of the generated models with convenience classes.
"""
from pydantic import BaseModel, ValidationInfo, BeforeValidator
from typing import Annotated, TypeVar, Type
from linkml.generators.pydanticgen.template import Imports, Import, ObjectImport
ModelType = TypeVar("ModelType", bound=Type[BaseModel])
# inspect.getsource() doesn't work for typevars because everything in the typing module
# doesn't behave like a normal python object
ModelTypeString = """ModelType = TypeVar("ModelType", bound=Type[BaseModel])"""
def _get_name(item: BaseModel | dict, info: ValidationInfo):
assert isinstance(item, (BaseModel, dict))
name = info.field_name
if isinstance(item, BaseModel):
item.name = name
else:
item['name'] = name
return item
Named = Annotated[ModelType, BeforeValidator(_get_name)]
"""
Generic annotated type that sets the ``name`` field of a model
to the name of the field with this type.
Examples:
class ChildModel(BaseModel):
name: str
value: int
class MyModel(BaseModel):
named_field: Named[ChildModel]
instance = MyModel(named_field={'value': 1})
instance.named_field.name == "named_field"
"""
NamedString = """Named = Annotated[ModelType, BeforeValidator(_get_name)]"""
NamedImports = Imports(imports=[
Import(module="typing", objects=[ObjectImport(name="Annotated"),ObjectImport(name="Type"), ObjectImport(name="TypeVar"), ]),
Import(module="pydantic", objects=[ObjectImport(name="ValidationInfo"), ObjectImport(name="BeforeValidator")])
])

View file

@ -120,7 +120,7 @@ def load_namespace_adapter(
return adapter return adapter
def load_nwb_core(core_version: str = "2.6.0", hdmf_version: str = "1.5.0") -> NamespacesAdapter: def load_nwb_core(core_version: str = "2.7.0", hdmf_version: str = "1.8.0") -> NamespacesAdapter:
""" """
Convenience function for loading the NWB core schema + hdmf-common as a namespace adapter. Convenience function for loading the NWB core schema + hdmf-common as a namespace adapter.

View file

@ -0,0 +1,17 @@
from pydantic import BaseModel
from nwb_linkml.includes import Named
def test_named_generic():
"""
the Named type should fill in the ``name`` field in a model from the field name
"""
class Child(BaseModel):
name: str
value: int
class Parent(BaseModel):
field_name: Named[Child]
# should instantiate correctly and have name set
instance = Parent(field_name={'value': 1})
assert instance.field_name.name == 'field_name'