diff --git a/nwb_linkml/src/nwb_linkml/providers/__init__.py b/nwb_linkml/src/nwb_linkml/providers/__init__.py index 175abce..700db3a 100644 --- a/nwb_linkml/src/nwb_linkml/providers/__init__.py +++ b/nwb_linkml/src/nwb_linkml/providers/__init__.py @@ -1,7 +1,27 @@ """ -Classes used for acquiring things from elsewhere, managing build processes, and caching results. +Classes used for acquiring things from elsewhere, managing build processes, and caching +results.[^recall] + +Each provider class handles gathering the inputs and caching and providing the desired outputs +for its domain. + +- {mod}`.git` providers faithfully yield to us the state of some code or schema somewhere +- {mod}`~.providers.linkml` providers nobly convert NWB schema language to linkml +- {mod}`~.providers.pydantic` providers courageously generate pydantic models from that linkml +- {mod}`~.providers.schema` providers sit around on their ass all day and wrap the other kinds of + providers for a simpler interface + +[^recall]: Recall that + - {mod}`.adapters` are for converting nwb schema language to linkml + - {mod}`.generators` (or rather the pydanticgen) are for converting linkml into concrete + representations like pydantic models + - so **providers** are the things that manage the multistage build process given some cache of + the stages so that you can go like "gimme model i want my model" and you will probably get it. """ +# ruff: noqa: I001 - import order necessary to avoid circular imports :) +from nwb_linkml.providers.provider import Provider +from nwb_linkml.providers.linkml import LinkMLProvider +from nwb_linkml.providers.pydantic import PydanticProvider +from nwb_linkml.providers.schema import SchemaProvider -from nwb_linkml.providers.schema import LinkMLProvider, PydanticProvider, SchemaProvider - -__all__ = ["LinkMLProvider", "PydanticProvider", "SchemaProvider"] +__all__ = ["Provider", "LinkMLProvider", "PydanticProvider", "SchemaProvider"] diff --git a/nwb_linkml/src/nwb_linkml/providers/linkml.py b/nwb_linkml/src/nwb_linkml/providers/linkml.py new file mode 100644 index 0000000..2c5db58 --- /dev/null +++ b/nwb_linkml/src/nwb_linkml/providers/linkml.py @@ -0,0 +1,282 @@ +import shutil +from pathlib import Path +from typing import Dict, Optional, TypedDict + +from linkml_runtime import SchemaView +from linkml_runtime.dumpers import yaml_dumper +from linkml_runtime.linkml_model import SchemaDefinition, SchemaDefinitionName + +from nwb_linkml import adapters, io +from nwb_linkml.adapters import BuildResult +from nwb_linkml.maps.naming import relative_path +from nwb_linkml.providers import Provider +from nwb_linkml.providers.git import DEFAULT_REPOS +from nwb_linkml.ui import AdapterProgress +from nwb_schema_language import Namespaces + + +class LinkMLSchemaBuild(TypedDict): + """Build result from :meth:`.LinkMLProvider.build`""" + + version: str + namespace: Path + name: str + result: Optional[BuildResult] + + +class LinkMLProvider(Provider): + """ + Provider for conversions from nwb schema language to linkML. + + By default, generate and manage a nest of temporary cache directories + (as configured by :class:`.Config`) for each version of a given namespace. + + Like other :class:`.Provider` classes, this model is not a singleton but + behaves a bit like one in that when instantiated without arguments + it is stateless (except for configuration by environment-level variables). + So we don't use ``@classmethod`` s here, but instantiating the class should remain + cheap. + + Namespaces can be built from: + + * namespace .yaml files: :meth:`.build_from_yaml` + * dictionaries, as are usually packaged in nwb files: :meth:`.build_from_dicts` + + All of which feed into... + + * :class:`~.adapters.NamespacesAdapter` used throughout the rest of ``nwb_linkml`` - + :meth:`.build` + + After a namespace is built, it can be accessed using :meth:`.LinkMLProvider.get`, which + can also be consumed by other providers, so a given namespace and version should only need + to be built once. + + Note: + At the moment there is no checking (eg. by comparing hashes) of different sources that + purport to be a given version of a namespace. When ambiguous, the class prefers to + build sets of namespaces together and use the most recently built ones since there is no + formal system for linking versions of namespaced schemas in nwb schema language. + + Examples: + + .. code-block:: python + + provider = LinkMLProvider() + # Simplest case, get the core nwb schema from the default NWB core repo + core = provider.get('core') + # Get a specific version of the core schema + core_other_version = provider.get('core', '2.2.0') + # Build a custom schema and then get it + # provider.build_from_yaml('myschema.yaml') + # my_schema = provider.get('myschema') + + """ + + PROVIDES = "linkml" + PROVIDES_CLASS = SchemaDefinition + + @property + def path(self) -> Path: + """``linkml_dir`` provided by :class:`.Config`""" + return self.config.linkml_dir + + def build_from_yaml( + self, path: Path, **kwargs: dict + ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: + """ + Build a namespace's schema + + Arguments: + path (:class:`pathlib.Path`): Path to the namespace .yaml + kwargs: passed to :meth:`.build` + """ + ns_adapter = adapters.NamespacesAdapter.from_yaml(path) + return self.build(ns_adapter, **kwargs) + + def build_from_dicts( + self, schemas: Dict[str, dict], **kwargs: dict + ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: + """ + Build from schema dictionaries, eg. as come from nwb files + + Arguments: + schemas (dict): A dictionary of ``{'schema_name': {:schema_definition}}``. + The "namespace" schema should have the key ``namespace``, which is used + to infer version and schema name. Post-load maps should have already + been applied + """ + ns_adapters = {} + for ns_name, ns_schemas in schemas.items(): + ns = Namespaces(**ns_schemas["namespace"]) + typed_schemas = [ + io.schema.load_schema_file(path=Path(key + ".yaml"), yaml=val) + for key, val in ns_schemas.items() + if key != "namespace" + ] + ns_adapter = adapters.NamespacesAdapter(namespaces=ns, schemas=typed_schemas) + ns_adapters[ns_name] = ns_adapter + + # get the correct imports + for adapter in ns_adapters.values(): + for schema_needs in adapter.needed_imports.values(): + for needed in schema_needs: + adapter.imported.append(ns_adapters[needed]) + + # then do the build + res = {} + for adapter in ns_adapters.values(): + res.update(self.build(adapter, **kwargs)) + + return res + + def build( + self, + ns_adapter: adapters.NamespacesAdapter, + versions: Optional[dict] = None, + dump: bool = True, + force: bool = False, + ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: + """ + Arguments: + namespaces (:class:`.NamespacesAdapter`): Adapter + (populated with any necessary imported namespaces) to build + versions (dict): Dict of specific versions to use + for cross-namespace imports. as ``{'namespace': 'version'}`` + If none is provided, use the most recent version + available. + dump (bool): If ``True`` (default), dump generated schema to YAML. otherwise just return + force (bool): If ``False`` (default), don't build schema that already exist. + If ``True`` , clear directory and rebuild + + Returns: + Dict[str, LinkMLSchemaBuild]. For normal builds, + :attr:`.LinkMLSchemaBuild.result` will be populated with results + of the build. If ``force == False`` and the schema already exist, it will be ``None`` + """ + + # Return cached result if available + if not force and all( + [ + (self.namespace_path(ns, version) / "namespace.yaml").exists() + for ns, version in ns_adapter.versions.items() + ] + ): + return { + k: LinkMLSchemaBuild( + name=k, + result=None, + namespace=self.namespace_path(k, v) / "namespace.yaml", + version=v, + ) + for k, v in ns_adapter.versions.items() + } + + if self.verbose: + progress = AdapterProgress(ns_adapter) + with progress: + built = ns_adapter.build(progress=progress) + else: + built = ns_adapter.build() + + # write schemas to yaml files + build_result = {} + + namespace_sch = [ + sch + for sch in built.schemas + if "is_namespace" in sch.annotations + and sch.annotations["is_namespace"].value in ("True", True) + ] + for ns_linkml in namespace_sch: + version = ns_adapter.versions[ns_linkml.name] + + # prepare the output directory + version_path = self.namespace_path(ns_linkml.name, version, allow_repo=False) + if version_path.exists() and force: + shutil.rmtree(str(version_path)) + version_path.mkdir(exist_ok=True, parents=True) + ns_file = version_path / "namespace.yaml" + + # schema built as part of this namespace that aren't the namespace file + other_schema = [ + sch + for sch in built.schemas + if sch.name.split(".")[0] == ns_linkml.name and sch not in namespace_sch + ] + + if force or not ns_file.exists(): + ns_linkml = self._fix_schema_imports(ns_linkml, ns_adapter, ns_file) + yaml_dumper.dump(ns_linkml, ns_file) + + # write the schemas for this namespace + for sch in other_schema: + output_file = version_path / (sch.name + ".yaml") + # fix the paths for intra-schema imports + sch = self._fix_schema_imports(sch, ns_adapter, output_file) + yaml_dumper.dump(sch, output_file) + + # make return result for just this namespace + build_result[ns_linkml.name] = LinkMLSchemaBuild( + namespace=ns_file, + name=ns_linkml.name, + result=BuildResult(schemas=[ns_linkml, *other_schema]), + version=version, + ) + + return build_result + + def _fix_schema_imports( + self, sch: SchemaDefinition, ns_adapter: adapters.NamespacesAdapter, output_file: Path + ) -> SchemaDefinition: + """ + Look there are a lot of nested directories around here and they aren't necessarily + all available at generation time. We have to relink relative imports between schema as + they are split and generated from the NWB. + """ + for animport in sch.imports: + if animport.split(".")[0] in ns_adapter.versions: + imported_path = ( + self.namespace_path( + animport.split(".")[0], ns_adapter.versions[animport.split(".")[0]] + ) + / "namespace" + ) + rel_path = relative_path(imported_path, output_file) + if str(rel_path) == "." or str(rel_path) == "namespace": + # same directory, just keep the existing import + continue + idx = sch.imports.index(animport) + del sch.imports[idx] + sch.imports.insert(idx, str(rel_path)) + return sch + + def get(self, namespace: str, version: Optional[str] = None) -> SchemaView: + """ + Get a schema view over the namespace. + + If a matching path for the namespace and version exists in the :attr:`.path`, + then return the SchemaView over that namespace. + + Otherwise, try and find a source using our :data:`.providers.git.DEFAULT_REPOS`. + + If none is found, then you need to build and cache the (probably custom) schema first with + :meth:`.build` + """ + path = self.namespace_path(namespace, version) / "namespace.yaml" + if not path.exists(): + path = self._find_source(namespace, version) + sv = SchemaView(path) + sv.path = path + return sv + + def _find_source(self, namespace: str, version: Optional[str] = None) -> Path: + """Try and find the namespace if it exists in our default repository and build it!""" + ns_repo = DEFAULT_REPOS.get(namespace, None) + if ns_repo is None: + raise KeyError( + f"Namespace {namespace} could not be found, and no git repository source has been" + " configured!" + ) + ns_file = ns_repo.provide_from_git(commit=version) + res = self.build_from_yaml(ns_file) + return res[namespace]["namespace"] diff --git a/nwb_linkml/src/nwb_linkml/providers/provider.py b/nwb_linkml/src/nwb_linkml/providers/provider.py new file mode 100644 index 0000000..6d01c3c --- /dev/null +++ b/nwb_linkml/src/nwb_linkml/providers/provider.py @@ -0,0 +1,163 @@ +import importlib +import os +from abc import ABC, abstractmethod +from pathlib import Path +from typing import Any, Dict, List, Optional, TypeVar + +from nwb_linkml import Config +from nwb_linkml.maps.naming import module_case, version_module_case + +P = TypeVar("P") + + +class Provider(ABC): + """ + Metaclass for different kind of providers! + + Args: + path (:class:`pathlib.Path`): Override the temporary directory configured by + the environment-wide :class:`.Config` object as the base directory that the + subclasses provide to. + verbose (bool): If ``True``, print things like progress bars to stdout :) + + Attributes: + config (:class:`.Config`): Configuration for the directories used by this + provider, unless overridden by ``path`` + allow_repo (bool): Allow the pathfinder to return the installed repository/package, + useful to enforce building into temporary directories, decoupling finding a path + during loading vs. building. Building into the repo is still possible if both + namespace and version are provided (ie. the path is fully qualified) and + :attr:`.config`'s path is the repository path. + cache_dir (:class:`pathlib.Path`): The main cache directory under which the other + providers will store the things they provide + """ + + PROVIDES: str + PROVIDES_CLASS: P = None + + def __init__(self, path: Optional[Path] = None, allow_repo: bool = True, verbose: bool = True): + config = Config(cache_dir=path) if path is not None else Config() + self.config = config + self.cache_dir = config.cache_dir + self.allow_repo = allow_repo + self.verbose = verbose + + @property + @abstractmethod + def path(self) -> Path: + """ + Base path for this kind of provider + """ + + @abstractmethod + def build(self, *args: Any) -> P: + """ + Whatever needs to be done to build this thing, if applicable + """ + + @abstractmethod + def get(self, *args: Any) -> Any: + """ + Get a cached item. + + Optionally, try any build it if it's possible to do so + """ + + def namespace_path( + self, namespace: str, version: Optional[str] = None, allow_repo: Optional[bool] = None + ) -> Path: + """ + Get the location for a given namespace of this type. + + Note that we don't check for existence, because this method should + also be used when generating schema --- this is the canonical location + + Arguments: + namespace (str): Namespace to get! + version (str): Optional, version of namespace. If ``None``, + either get the most recent version built, or if + ``namespace`` is ``core`` or ``hdmf-common``, use the + modules provided with the package. We do not use the most + recent *version*, but the most recently *generated* version + because it's assumed that's the one you want if you're just + gesturally reaching for one. + allow_repo (bool): Optional - override instance-level ``allow_repo`` attr + """ + if allow_repo is None: + allow_repo = self.allow_repo + + namespace_module = module_case(namespace) + namespace_path = self.path / namespace_module + if not namespace_path.exists() and namespace in ("core", "hdmf-common") and allow_repo: + # return builtins + module_path = Path(importlib.util.find_spec("nwb_linkml").origin).parent + + if self.PROVIDES == "linkml": + namespace_path = module_path / "schema" / "linkml" / namespace + elif self.PROVIDES == "pydantic": + namespace_path = module_path / "models" / "pydantic" / namespace + + if version is not None: + version_path = namespace_path / version_module_case(version) + # version_path.mkdir(exist_ok=True, parents=True) + else: + # or find the most recently built one + versions = sorted(namespace_path.iterdir(), key=os.path.getmtime) + versions = [v for v in versions if v.is_dir() and v.name not in ("__pycache__")] + if len(versions) == 0: + raise FileNotFoundError("No version provided, and no existing schema found") + version_path = versions[-1] + + return version_path + + @property + def available_versions(self) -> Dict[str, List[str]]: + """ + Dictionary mapping a namespace to a list of built versions + """ + from nwb_linkml.providers import LinkMLProvider + versions = {} # type: Dict[str, List[Path]] + + # first get any builtins provided by the package itself + # these get overwritten by + module_path = Path(importlib.util.find_spec("nwb_linkml").origin).parent + builtin_namespaces = [] + if self.PROVIDES == "linkml": + namespace_path = module_path / "schema" / "linkml" + builtin_namespaces = list(namespace_path.iterdir()) + elif self.PROVIDES == "pydantic": + namespace_path = module_path / "models" / "pydantic" + builtin_namespaces = list(namespace_path.iterdir()) + + tmp_paths = [] + try: + tmp_paths.extend(list(self.path.iterdir())) + if self.PROVIDES == "pydantic": + # we also include versions that we just have the linkml version of + # because they are also available, we just have to build them. + # maybe the semantics of this property are getting overloaded tho + # and we need to separate the maintenance of the temporary directory + # from providing from it + tmp_paths.extend(list(LinkMLProvider(path=self.config.cache_dir).path.iterdir())) + except FileNotFoundError: + # fine, just return the builtins + pass + + for ns_dir in builtin_namespaces + tmp_paths: + if not ns_dir.is_dir(): + continue + if ns_dir.name not in versions: + versions[ns_dir.name] = [] + + versions[ns_dir.name].extend([v for v in ns_dir.iterdir() if v.is_dir()]) + + # flatten out in case we got duplicates between the builtins and cache + res = { + k: [ + v.name for v in sorted(set(v_paths), key=os.path.getmtime) if v.name.startswith("v") + ] + for k, v_paths in versions.items() + } + return res + + diff --git a/nwb_linkml/src/nwb_linkml/providers/pydantic.py b/nwb_linkml/src/nwb_linkml/providers/pydantic.py new file mode 100644 index 0000000..c6a2130 --- /dev/null +++ b/nwb_linkml/src/nwb_linkml/providers/pydantic.py @@ -0,0 +1,449 @@ +""" +Provider for pydantic models. +""" +import importlib +import sys +from importlib.abc import MetaPathFinder +from importlib.machinery import ModuleSpec +from pathlib import Path +from types import ModuleType +from typing import List, Optional, Type + +from pydantic import BaseModel + +from nwb_linkml import io +from nwb_linkml.generators.pydantic import NWBPydanticGenerator +from nwb_linkml.maps.naming import module_case, version_module_case +from nwb_linkml.providers import LinkMLProvider, Provider + + +class PydanticProvider(Provider): + """ + Provider for pydantic models built from linkml-style nwb schema + (ie. as provided by :class:`.LinkMLProvider`) + + Generates pydantic models into a :attr:`~.PydanticProvider.path` that can be imported + as if they were within the `nwb_linkml` namespace using an :class:`.EctopicModelFinder` . + + .. todo:: + + Documentation of directory structure and caching will be completed once it is finalized :) + + """ + + PROVIDES = "pydantic" + + def __init__(self, path: Optional[Path] = None, verbose: bool = True): + super().__init__(path, verbose) + # create a metapathfinder to find module we might create + pathfinder = EctopicModelFinder(self.path) + sys.meta_path.append(pathfinder) + + @property + def path(self) -> Path: + """``pydantic_dir`` provided by :class:`.Config`""" + return self.config.pydantic_dir + + def build( + self, + namespace: str | Path, + out_file: Optional[Path] = None, + version: Optional[str] = None, + versions: Optional[dict] = None, + split: bool = True, + dump: bool = True, + force: bool = False, + **kwargs: dict, + ) -> str | List[str]: + """ + + Notes: + We currently infer namespace and version from the path when ``namespace`` is a Path, + which is a patently Bad Thing To Do. This is a temporary measure until we decide on + a permanent means by which we want to cache built artifacts <3. Hierarchies of folders + is not the target design. + + + Args: + namespace (Union[str, :class:`pathlib.Path`]): If a string, use a + :class:`.LinkMLProvider` to get the converted schema. If a path, + assume we have been given an explicit ``namespace.yaml`` from a converted + NWB -> LinkML schema to load from. + out_file (Optional[Path]): Optionally override the output file. If ``None``, + generate from namespace and version + version (Optional[str]): The version of the schema to build, if present. + Works similarly to ``version`` in :class:`.LinkMLProvider`. + Ignored if ``namespace`` is a Path. + versions (Optional[dict]): An explicit mapping of namespaces and versions to use when + building the combined pydantic `namespace.py` file. + Since NWB doesn't have an explicit version dependency system between schema, + there is intrinsic ambiguity between which version + of which schema should be used when imported from another. + This mapping allows those ambiguities to be resolved. + See :class:`.NWBPydanticGenerator` 's ``versions`` argument for more information. + split (bool): If ``False`` (default), generate a single ``namespace.py`` file, + otherwise generate a python file for each schema in the namespace + in addition to a ``namespace.py`` that imports from them + dump (bool): If ``True`` (default), dump the model to the cache, + otherwise just return the serialized string of built pydantic model + force (bool): If ``False`` (default), don't build the model if it already exists, + if ``True`` , delete and rebuild any model + **kwargs: Passed to :class:`.NWBPydanticGenerator` + + Returns: + str: The built model file as returned from :meth:`.NWBPydanticGenerator.serialize` + """ + + if isinstance(namespace, str) and not ( + namespace.endswith(".yaml") or namespace.endswith(".yml") + ): + # we're given a name of a namespace to build + name = namespace + path = ( + LinkMLProvider(path=self.config.cache_dir).namespace_path(namespace, version) + / "namespace.yaml" + ) + if version is None: + # Get the most recently built version + version = LinkMLProvider(path=self.config.cache_dir).available_versions[name][-1] + fn = path.parts[-1] + else: + # given a path to a namespace linkml yaml file + path = Path(namespace) + # FIXME: this is extremely fragile, but get the details from the path. + # this is faster than reading yaml for now + name = path.parts[-3] + version = path.parts[-2] + fn = path.parts[-1] + + version = version_module_case(version) + # this is extremely fragile, we should not be inferring version number from paths... + # TODO: we need an efficient peek for specific keys within a yaml file + if out_file is None: + fn = fn.removesuffix(".yaml") + fn = module_case(fn) + ".py" + out_file = self.path / name / version / fn + + default_kwargs = { + "split": split, + "emit_metadata": True, + "gen_slots": True, + "pydantic_version": "2", + } + default_kwargs.update(kwargs) + + # if we weren't given explicit versions to load, figure them out from the namespace + if versions is None: + versions = self._get_dependent_versions(path) + + if split: + return self._build_split(path, versions, default_kwargs, dump, out_file, force) + else: + return self._build_unsplit(path, versions, default_kwargs, dump, out_file, force) + + def _build_unsplit( + self, + path: Path, + versions: dict, + default_kwargs: dict, + dump: bool, + out_file: Path, + force: bool, + ) -> Optional[str]: + if out_file.exists() and not force: + with open(out_file) as ofile: + serialized = ofile.read() + return serialized + + generator = NWBPydanticGenerator(str(path), versions=versions, **default_kwargs) + serialized = generator.serialize() + if dump: + out_file.parent.mkdir(parents=True, exist_ok=True) + with open(out_file, "w") as ofile: + ofile.write(serialized) + + self._make_inits(out_file) + + return serialized + + def _build_split( + self, + path: Path, + versions: dict, + default_kwargs: dict, + dump: bool, + out_file: Path, + force: bool, + ) -> List[str]: + serialized = [] + for schema_file in path.parent.glob("*.yaml"): + this_out = out_file.parent / (module_case(schema_file.stem) + ".py") + serialized.append( + self._build_unsplit(schema_file, versions, default_kwargs, dump, this_out, force) + ) + + # If there are dependent versions that also need to be built, do that now! + needed = [ + (module_case(ns), version_module_case(version)) + for ns, version in versions.items() + if version_module_case(version) not in self.available_versions.get(ns, []) + ] + for needs in needed: + needed_path = path.parents[2] / needs[0] / needs[1] / "namespace.yaml" + out_file_stem = out_file.parents[2] / needs[0] / needs[1] + for schema_file in needed_path.parent.glob("*.yaml"): + this_out = out_file_stem / (module_case(schema_file.stem) + ".py") + serialized.append( + self._build_unsplit( + schema_file, versions, default_kwargs, dump, this_out, force + ) + ) + + return serialized + + def _make_inits(self, out_file: Path): + """ + Make __init__.py files for the directory a model is output to and its immediate parent. + (ig this could be generalized in depth but lets not go nuts in here with the nesting) + """ + initfile = out_file.parent / "__init__.py" + parent_init = out_file.parent.parent / "__init__.py" + for ifile in (initfile, parent_init): + if not ifile.exists(): + with open(ifile, "w") as ifile_open: + ifile_open.write(" ") + + def _get_dependent_versions(self, path: Path) -> dict[str, str]: + """ + For a given namespace schema file, get the versions of any other schemas it imports + + Namespace imports will be in the importing schema like: + + imports: + -../../hdmf_common/v1_8_0/namespace + -../../hdmf_experimental/v0_5_0/namespace + + Returns: + dict[str,str]: A dictionary mapping a namespace to a version number + """ + schema = io.schema.load_yaml(path) + versions = {} + for i in schema["imports"]: + if i.startswith(".."): + import_path = (Path(path).parent / Path(i + ".yaml")).resolve() + imported_schema = io.schema.load_yaml(import_path) + versions[imported_schema["name"]] = imported_schema["version"] + return versions + + @classmethod + def module_name(self, namespace: str, version: str) -> str: + """Module name for the built module + + e.g.:: + + nwb_linkml.models.pydantic.{namespace}.{version} + """ + name_pieces = [ + "nwb_linkml", + "models", + "pydantic", + module_case(namespace), + version_module_case(version), + ] + module_name = ".".join(name_pieces) + return module_name + + def import_module(self, namespace: str, version: Optional[str] = None) -> ModuleType: + """ + Import a module within the temporary directory from its namespace and version + + In most cases, you're looking for :meth:`.PydanticProvider.get`, this method is + made available in case you don't want to accidentally build something + or invoke the rest of the provisioning system. + + Args: + namespace (str): Name of namespace + version (Optional[str]): Version to import, if None, + try and get the most recently built version. + + Returns: + :class:`types.ModuleType` + """ + # get latest version if None + if version is None: + version = self.available_versions[namespace][-1] + + path = self.namespace_path(namespace, version) / "namespace.py" + if not path.exists(): + raise ImportError(f"Module has not been built yet {path}") + module_name = self.module_name(namespace, version) + + # import module level first - when python does relative imports, + # it needs to have the parent modules imported separately + # this breaks split model creation when they are outside of the + # package repository (ie. loaded from an nwb file) because it tries + # to look for the containing namespace folder within the nwb_linkml package and fails + init_spec = importlib.util.spec_from_file_location(module_name, path.parent / "__init__.py") + init_module = importlib.util.module_from_spec(init_spec) + sys.modules[module_name] = init_module + init_spec.loader.exec_module(init_module) + + # then the namespace package + module_name = module_name + ".namespace" + spec = importlib.util.spec_from_file_location(module_name, path) + module = importlib.util.module_from_spec(spec) + sys.modules[module_name] = module + spec.loader.exec_module(module) + return module + + def get( + self, namespace: str, version: Optional[str] = None, allow_repo: Optional[bool] = None + ) -> ModuleType: + """ + Get the imported module for a given namespace and version. + + A given namespace will be stored in :data:`sys.modules` as + ``nwb_linkml.models.{namespace}``, + so first check if there is any already-imported module, and return that if so. + + Then we check in the temporary directory for an already-built ``namespace.py`` file + + Otherwise we pass arguments to :meth:`.PydanticProvider.build` and attempt to build them + before returning. + + Notes: + The imported modules shadow the "actual" + ``nwb_linkml.models`` module as would be imported from the usual location + within the package directory. + This is intentional, as models can then be used as if they were + integrated parts of the package, + and also so the active version of a namespace can be cleanly accessed + (ie. without ``from nwb_linkml.models.core import v2_2_0 as core`` ). + Accordingly, we assume that people will only be using a single version of NWB in a given + Python session. + + + Args: + namespace (str): Name of namespace to import. Must have either been previously built + with :meth:`.PydanticProvider.build` or + a matching namespace/version combo must be available to the + :class:`.LinkMLProvider` + version (Optional[str]): Version to import. If ``None``, + get the most recently build module + allow_repo (bool): Allow getting modules provided within + :mod:`nwb_linkml.models.pydantic` + + Returns: + The imported :class:`types.ModuleType` object that has all the built + classes at the root level. + + """ + if allow_repo is None: + allow_repo = self.allow_repo + + if version is None: + version = self.available_versions[namespace][-1] + + module_name = self.module_name(namespace, version) + namespace_name = module_name + ".namespace" + + if not allow_repo: + self._clear_package_imports() + + if namespace_name in sys.modules: + return sys.modules[namespace_name] + + try: + path = self.namespace_path(namespace, version, allow_repo) + except FileNotFoundError: + path = None + + if path is None or not path.exists(): + _ = self.build(namespace, version=version) + + module = self.import_module(namespace, version) + return module + + @staticmethod + def _clear_package_imports() -> None: + """ + When using allow_repo=False, delete any already-imported + namespaces from sys.modules that are within the nwb_linkml package + """ + # make sure we don't have any in-repo modules + repo_base = Path(__file__).parents[1] + deletes = [] + for k, v in sys.modules.items(): + if not k.startswith("nwb_linkml.models.pydantic"): + continue + try: + Path(v.__file__).relative_to(repo_base) + # if it is a subpath, delete it + deletes.append(k) + except ValueError: + # path is not a subpath + continue + for d in deletes: + del sys.modules[d] + + def get_class( + self, namespace: str, class_: str, version: Optional[str] = None + ) -> Type[BaseModel]: + """ + Get a class from a given namespace and version! + + Args: + namespace (str): Name of a namespace that has been previously built and cached, + otherwise we will attempt to build it from the + :data:`.providers.git.DEFAULT_REPOS` + class_ (str): Name of class to retrieve + version (Optional[str]): Optional version of the schema to retrieve from + + Returns: + :class:`pydantic.BaseModel` + """ + mod = self.get(namespace, version) + return getattr(mod, class_) + + +class EctopicModelFinder(MetaPathFinder): + """ + A meta path finder that allows the import machinery to find a model + package even if it might be outside the actual nwb_linkml namespace, + as occurs when building split models in a temporary directory. + + References: + - https://docs.python.org/3/reference/import.html#the-meta-path + - https://docs.python.org/3/library/importlib.html#importlib.abc.MetaPathFinder + + """ + + MODEL_STEM = "nwb_linkml.models.pydantic" + + def __init__(self, path: Path, *args, **kwargs): + super().__init__(*args, **kwargs) + self.path = path + + def find_spec( + self, fullname: str, path: Optional[str], target: Optional[ModuleType] = None + ) -> Optional[ModuleSpec]: + """If we are loading a generated pydantic module, return an importlib spec""" + if not fullname.startswith(self.MODEL_STEM): + return None + else: + # get submodule beneath model stem + submod = fullname.replace(self.MODEL_STEM, "").lstrip(".") + base_path = Path(self.path, *submod.split(".")) + + # switch if we're asked for a package or a module + mod_path = Path(str(base_path) + ".py") + pkg_path = base_path / "__init__.py" + if mod_path.exists(): + import_path = mod_path + elif pkg_path.exists(): + import_path = pkg_path + else: + return None + + spec = importlib.util.spec_from_file_location(fullname, import_path) + return spec diff --git a/nwb_linkml/src/nwb_linkml/providers/schema.py b/nwb_linkml/src/nwb_linkml/providers/schema.py index 66e73ab..adadb00 100644 --- a/nwb_linkml/src/nwb_linkml/providers/schema.py +++ b/nwb_linkml/src/nwb_linkml/providers/schema.py @@ -42,860 +42,14 @@ so eg. for the linkML and pydantic providers: """ -import importlib -import os -import shutil -import sys -from abc import ABC, abstractmethod -from importlib.abc import MetaPathFinder -from importlib.machinery import ModuleSpec from pathlib import Path from types import ModuleType -from typing import Any, Dict, List, Optional, Type, TypedDict, TypeVar +from typing import Dict, Optional, Type -from linkml_runtime import SchemaView -from linkml_runtime.dumpers import yaml_dumper -from linkml_runtime.linkml_model import SchemaDefinition, SchemaDefinitionName from pydantic import BaseModel -from nwb_linkml import adapters, io -from nwb_linkml.adapters.adapter import BuildResult -from nwb_linkml.config import Config -from nwb_linkml.generators.pydantic import NWBPydanticGenerator -from nwb_linkml.maps.naming import module_case, relative_path, version_module_case -from nwb_linkml.providers.git import DEFAULT_REPOS -from nwb_linkml.ui import AdapterProgress -from nwb_schema_language import Namespaces - -P = TypeVar("P") - - -class Provider(ABC): - """ - Metaclass for different kind of providers! - - Args: - path (:class:`pathlib.Path`): Override the temporary directory configured by - the environment-wide :class:`.Config` object as the base directory that the - subclasses provide to. - verbose (bool): If ``True``, print things like progress bars to stdout :) - - Attributes: - config (:class:`.Config`): Configuration for the directories used by this - provider, unless overridden by ``path`` - allow_repo (bool): Allow the pathfinder to return the installed repository/package, - useful to enforce building into temporary directories, decoupling finding a path - during loading vs. building. Building into the repo is still possible if both - namespace and version are provided (ie. the path is fully qualified) and - :attr:`.config`'s path is the repository path. - cache_dir (:class:`pathlib.Path`): The main cache directory under which the other - providers will store the things they provide - """ - - PROVIDES: str - PROVIDES_CLASS: P = None - - def __init__(self, path: Optional[Path] = None, allow_repo: bool = True, verbose: bool = True): - config = Config(cache_dir=path) if path is not None else Config() - self.config = config - self.cache_dir = config.cache_dir - self.allow_repo = allow_repo - self.verbose = verbose - - @property - @abstractmethod - def path(self) -> Path: - """ - Base path for this kind of provider - """ - - @abstractmethod - def build(self, *args: Any) -> P: - """ - Whatever needs to be done to build this thing, if applicable - """ - - @abstractmethod - def get(self, *args: Any) -> Any: - """ - Get a cached item. - - Optionally, try any build it if it's possible to do so - """ - - def namespace_path( - self, namespace: str, version: Optional[str] = None, allow_repo: Optional[bool] = None - ) -> Path: - """ - Get the location for a given namespace of this type. - - Note that we don't check for existence, because this method should - also be used when generating schema --- this is the canonical location - - Arguments: - namespace (str): Namespace to get! - version (str): Optional, version of namespace. If ``None``, - either get the most recent version built, or if - ``namespace`` is ``core`` or ``hdmf-common``, use the - modules provided with the package. We do not use the most - recent *version*, but the most recently *generated* version - because it's assumed that's the one you want if you're just - gesturally reaching for one. - allow_repo (bool): Optional - override instance-level ``allow_repo`` attr - """ - if allow_repo is None: - allow_repo = self.allow_repo - - namespace_module = module_case(namespace) - namespace_path = self.path / namespace_module - if not namespace_path.exists() and namespace in ("core", "hdmf-common") and allow_repo: - # return builtins - module_path = Path(importlib.util.find_spec("nwb_linkml").origin).parent - - if self.PROVIDES == "linkml": - namespace_path = module_path / "schema" / "linkml" / namespace - elif self.PROVIDES == "pydantic": - namespace_path = module_path / "models" / "pydantic" / namespace - - if version is not None: - version_path = namespace_path / version_module_case(version) - # version_path.mkdir(exist_ok=True, parents=True) - else: - # or find the most recently built one - versions = sorted(namespace_path.iterdir(), key=os.path.getmtime) - versions = [v for v in versions if v.is_dir() and v.name not in ("__pycache__")] - if len(versions) == 0: - raise FileNotFoundError("No version provided, and no existing schema found") - version_path = versions[-1] - - return version_path - - @property - def available_versions(self) -> Dict[str, List[str]]: - """ - Dictionary mapping a namespace to a list of built versions - """ - versions = {} # type: Dict[str, List[Path]] - - # first get any builtins provided by the package itself - # these get overwritten by - module_path = Path(importlib.util.find_spec("nwb_linkml").origin).parent - builtin_namespaces = [] - if self.PROVIDES == "linkml": - namespace_path = module_path / "schema" / "linkml" - builtin_namespaces = list(namespace_path.iterdir()) - elif self.PROVIDES == "pydantic": - namespace_path = module_path / "models" / "pydantic" - builtin_namespaces = list(namespace_path.iterdir()) - - tmp_paths = [] - try: - tmp_paths.extend(list(self.path.iterdir())) - if self.PROVIDES == "pydantic": - # we also include versions that we just have the linkml version of - # because they are also available, we just have to build them. - # maybe the semantics of this property are getting overloaded tho - # and we need to separate the maintenance of the temporary directory - # from providing from it - tmp_paths.extend(list(LinkMLProvider(path=self.config.cache_dir).path.iterdir())) - except FileNotFoundError: - # fine, just return the builtins - pass - - for ns_dir in builtin_namespaces + tmp_paths: - if not ns_dir.is_dir(): - continue - if ns_dir.name not in versions: - versions[ns_dir.name] = [] - - versions[ns_dir.name].extend([v for v in ns_dir.iterdir() if v.is_dir()]) - - # flatten out in case we got duplicates between the builtins and cache - res = { - k: [ - v.name for v in sorted(set(v_paths), key=os.path.getmtime) if v.name.startswith("v") - ] - for k, v_paths in versions.items() - } - return res - - -class LinkMLSchemaBuild(TypedDict): - """Build result from :meth:`.LinkMLProvider.build`""" - - version: str - namespace: Path - name: str - result: Optional[BuildResult] - - -class LinkMLProvider(Provider): - """ - Provider for conversions from nwb schema language to linkML. - - By default, generate and manage a nest of temporary cache directories - (as configured by :class:`.Config`) for each version of a given namespace. - - Like other :class:`.Provider` classes, this model is not a singleton but - behaves a bit like one in that when instantiated without arguments - it is stateless (except for configuration by environment-level variables). - So we don't use ``@classmethod`` s here, but instantiating the class should remain - cheap. - - Namespaces can be built from: - - * namespace .yaml files: :meth:`.build_from_yaml` - * dictionaries, as are usually packaged in nwb files: :meth:`.build_from_dicts` - - All of which feed into... - - * :class:`~.adapters.NamespacesAdapter` used throughout the rest of ``nwb_linkml`` - - :meth:`.build` - - After a namespace is built, it can be accessed using :meth:`.LinkMLProvider.get`, which - can also be consumed by other providers, so a given namespace and version should only need - to be built once. - - Note: - At the moment there is no checking (eg. by comparing hashes) of different sources that - purport to be a given version of a namespace. When ambiguous, the class prefers to - build sets of namespaces together and use the most recently built ones since there is no - formal system for linking versions of namespaced schemas in nwb schema language. - - Examples: - - .. code-block:: python - - provider = LinkMLProvider() - # Simplest case, get the core nwb schema from the default NWB core repo - core = provider.get('core') - # Get a specific version of the core schema - core_other_version = provider.get('core', '2.2.0') - # Build a custom schema and then get it - # provider.build_from_yaml('myschema.yaml') - # my_schema = provider.get('myschema') - - """ - - PROVIDES = "linkml" - PROVIDES_CLASS = SchemaDefinition - - @property - def path(self) -> Path: - """``linkml_dir`` provided by :class:`.Config`""" - return self.config.linkml_dir - - def build_from_yaml( - self, path: Path, **kwargs: dict - ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: - """ - Build a namespace's schema - - Arguments: - path (:class:`pathlib.Path`): Path to the namespace .yaml - kwargs: passed to :meth:`.build` - """ - ns_adapter = adapters.NamespacesAdapter.from_yaml(path) - return self.build(ns_adapter, **kwargs) - - def build_from_dicts( - self, schemas: Dict[str, dict], **kwargs: dict - ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: - """ - Build from schema dictionaries, eg. as come from nwb files - - Arguments: - schemas (dict): A dictionary of ``{'schema_name': {:schema_definition}}``. - The "namespace" schema should have the key ``namespace``, which is used - to infer version and schema name. Post-load maps should have already - been applied - """ - ns_adapters = {} - for ns_name, ns_schemas in schemas.items(): - ns = Namespaces(**ns_schemas["namespace"]) - typed_schemas = [ - io.schema.load_schema_file(path=Path(key + ".yaml"), yaml=val) - for key, val in ns_schemas.items() - if key != "namespace" - ] - ns_adapter = adapters.NamespacesAdapter(namespaces=ns, schemas=typed_schemas) - ns_adapters[ns_name] = ns_adapter - - # get the correct imports - for adapter in ns_adapters.values(): - for schema_needs in adapter.needed_imports.values(): - for needed in schema_needs: - adapter.imported.append(ns_adapters[needed]) - - # then do the build - res = {} - for adapter in ns_adapters.values(): - res.update(self.build(adapter, **kwargs)) - - return res - - def build( - self, - ns_adapter: adapters.NamespacesAdapter, - versions: Optional[dict] = None, - dump: bool = True, - force: bool = False, - ) -> Dict[str | SchemaDefinitionName, LinkMLSchemaBuild]: - """ - Arguments: - namespaces (:class:`.NamespacesAdapter`): Adapter - (populated with any necessary imported namespaces) to build - versions (dict): Dict of specific versions to use - for cross-namespace imports. as ``{'namespace': 'version'}`` - If none is provided, use the most recent version - available. - dump (bool): If ``True`` (default), dump generated schema to YAML. otherwise just return - force (bool): If ``False`` (default), don't build schema that already exist. - If ``True`` , clear directory and rebuild - - Returns: - Dict[str, LinkMLSchemaBuild]. For normal builds, - :attr:`.LinkMLSchemaBuild.result` will be populated with results - of the build. If ``force == False`` and the schema already exist, it will be ``None`` - """ - - if not force and all( - [ - (self.namespace_path(ns, version) / "namespace.yaml").exists() - for ns, version in ns_adapter.versions.items() - ] - ): - return { - k: LinkMLSchemaBuild( - name=k, - result=None, - namespace=self.namespace_path(k, v) / "namespace.yaml", - version=v, - ) - for k, v in ns_adapter.versions.items() - } - - # self._find_imports(ns_adapter, versions, populate=True) - if self.verbose: - progress = AdapterProgress(ns_adapter) - # progress.start() - with progress: - built = ns_adapter.build(progress=progress) - else: - progress = None - built = ns_adapter.build() - - # write schemas to yaml files - build_result = {} - - namespace_sch = [ - sch - for sch in built.schemas - if "is_namespace" in sch.annotations - and sch.annotations["is_namespace"].value in ("True", True) - ] - for ns_linkml in namespace_sch: - version = ns_adapter.versions[ns_linkml.name] - version_path = self.namespace_path(ns_linkml.name, version, allow_repo=False) - if version_path.exists() and force: - shutil.rmtree(str(version_path)) - version_path.mkdir(exist_ok=True, parents=True) - ns_file = version_path / "namespace.yaml" - # schema built as part of this namespace that aren't the namespace file - other_schema = [ - sch - for sch in built.schemas - if sch.name.split(".")[0] == ns_linkml.name and sch not in namespace_sch - ] - - if force or not ns_file.exists(): - ns_linkml = self._fix_schema_imports(ns_linkml, ns_adapter, ns_file) - yaml_dumper.dump(ns_linkml, ns_file) - - # write the schemas for this namespace - for sch in other_schema: - output_file = version_path / (sch.name + ".yaml") - # fix the paths for intra-schema imports - sch = self._fix_schema_imports(sch, ns_adapter, output_file) - yaml_dumper.dump(sch, output_file) - - # make return result for just this namespace - build_result[ns_linkml.name] = LinkMLSchemaBuild( - namespace=ns_file, - name=ns_linkml.name, - result=BuildResult(schemas=[ns_linkml, *other_schema]), - version=version, - ) - - return build_result - - def _fix_schema_imports( - self, sch: SchemaDefinition, ns_adapter: adapters.NamespacesAdapter, output_file: Path - ) -> SchemaDefinition: - for animport in sch.imports: - if animport.split(".")[0] in ns_adapter.versions: - imported_path = ( - self.namespace_path( - animport.split(".")[0], ns_adapter.versions[animport.split(".")[0]] - ) - / "namespace" - ) - rel_path = relative_path(imported_path, output_file) - if str(rel_path) == "." or str(rel_path) == "namespace": - # same directory, just keep the existing import - continue - idx = sch.imports.index(animport) - del sch.imports[idx] - sch.imports.insert(idx, str(rel_path)) - return sch - - def get(self, namespace: str, version: Optional[str] = None) -> SchemaView: - """ - Get a schema view over the namespace. - - If a matching path for the namespace and version exists in the :attr:`.path`, - then return the SchemaView over that namespace. - - Otherwise, try and find a source using our :data:`.providers.git.DEFAULT_REPOS`. - - If none is found, then you need to build and cache the (probably custom) schema first with - :meth:`.build` - """ - path = self.namespace_path(namespace, version) / "namespace.yaml" - if not path.exists(): - path = self._find_source(namespace, version) - sv = SchemaView(path) - sv.path = path - return sv - - def _find_source(self, namespace: str, version: Optional[str] = None) -> Path: - """Try and find the namespace if it exists in our default repository and build it!""" - ns_repo = DEFAULT_REPOS.get(namespace, None) - if ns_repo is None: - raise KeyError( - f"Namespace {namespace} could not be found, and no git repository source has been" - " configured!" - ) - ns_file = ns_repo.provide_from_git(commit=version) - res = self.build_from_yaml(ns_file) - return res[namespace]["namespace"] - - -class PydanticProvider(Provider): - """ - Provider for pydantic models built from linkml-style nwb schema - (ie. as provided by :class:`.LinkMLProvider`) - """ - - PROVIDES = "pydantic" - - def __init__(self, path: Optional[Path] = None, verbose: bool = True): - super().__init__(path, verbose) - # create a metapathfinder to find module we might create - pathfinder = EctopicModelFinder(self.path) - sys.meta_path.append(pathfinder) - - @property - def path(self) -> Path: - """``pydantic_dir`` provided by :class:`.Config`""" - return self.config.pydantic_dir - - def build( - self, - namespace: str | Path, - out_file: Optional[Path] = None, - version: Optional[str] = None, - versions: Optional[dict] = None, - split: bool = True, - dump: bool = True, - force: bool = False, - **kwargs: dict, - ) -> str | List[str]: - """ - - Notes: - We currently infer namespace and version from the path when ``namespace`` is a Path, - which is a patently Bad Thing To Do. This is a temporary measure until we decide on - a permanent means by which we want to cache built artifacts <3. Hierarchies of folders - is not the target design. - - - Args: - namespace (Union[str, :class:`pathlib.Path`]): If a string, use a - :class:`.LinkMLProvider` to get the converted schema. If a path, - assume we have been given an explicit ``namespace.yaml`` from a converted - NWB -> LinkML schema to load from. - out_file (Optional[Path]): Optionally override the output file. If ``None``, - generate from namespace and version - version (Optional[str]): The version of the schema to build, if present. - Works similarly to ``version`` in :class:`.LinkMLProvider`. - Ignored if ``namespace`` is a Path. - versions (Optional[dict]): An explicit mapping of namespaces and versions to use when - building the combined pydantic `namespace.py` file. - Since NWB doesn't have an explicit version dependency system between schema, - there is intrinsic ambiguity between which version - of which schema should be used when imported from another. - This mapping allows those ambiguities to be resolved. - See :class:`.NWBPydanticGenerator` 's ``versions`` argument for more information. - split (bool): If ``False`` (default), generate a single ``namespace.py`` file, - otherwise generate a python file for each schema in the namespace - in addition to a ``namespace.py`` that imports from them - dump (bool): If ``True`` (default), dump the model to the cache, - otherwise just return the serialized string of built pydantic model - force (bool): If ``False`` (default), don't build the model if it already exists, - if ``True`` , delete and rebuild any model - **kwargs: Passed to :class:`.NWBPydanticGenerator` - - Returns: - str: The built model file as returned from :meth:`.NWBPydanticGenerator.serialize` - """ - - if isinstance(namespace, str) and not ( - namespace.endswith(".yaml") or namespace.endswith(".yml") - ): - # we're given a name of a namespace to build - name = namespace - path = ( - LinkMLProvider(path=self.config.cache_dir).namespace_path(namespace, version) - / "namespace.yaml" - ) - if version is None: - # Get the most recently built version - version = LinkMLProvider(path=self.config.cache_dir).available_versions[name][-1] - fn = path.parts[-1] - else: - # given a path to a namespace linkml yaml file - path = Path(namespace) - # FIXME: this is extremely fragile, but get the details from the path. - # this is faster than reading yaml for now - name = path.parts[-3] - version = path.parts[-2] - fn = path.parts[-1] - - version = version_module_case(version) - # this is extremely fragile, we should not be inferring version number from paths... - if out_file is None: - fn = fn.removesuffix(".yaml") - fn = module_case(fn) + ".py" - out_file = self.path / name / version / fn - - default_kwargs = { - "split": split, - "emit_metadata": True, - "gen_slots": True, - "pydantic_version": "2", - } - default_kwargs.update(kwargs) - - # if we weren't given explicit versions to load, figure them out from the namespace - if versions is None: - versions = self._get_dependent_versions(path) - - if split: - return self._build_split(path, versions, default_kwargs, dump, out_file, force) - else: - return self._build_unsplit(path, versions, default_kwargs, dump, out_file, force) - - def _build_unsplit( - self, - path: Path, - versions: dict, - default_kwargs: dict, - dump: bool, - out_file: Path, - force: bool, - ) -> Optional[str]: - if out_file.exists() and not force: - with open(out_file) as ofile: - serialized = ofile.read() - return serialized - - generator = NWBPydanticGenerator(str(path), versions=versions, **default_kwargs) - serialized = generator.serialize() - if dump: - out_file.parent.mkdir(parents=True, exist_ok=True) - with open(out_file, "w") as ofile: - ofile.write(serialized) - - # make initfiles for this directory and parent, - initfile = out_file.parent / "__init__.py" - parent_init = out_file.parent.parent / "__init__.py" - for ifile in (initfile, parent_init): - if not ifile.exists(): - with open(ifile, "w") as ifile_open: - ifile_open.write(" ") - - return serialized - - def _build_split( - self, - path: Path, - versions: dict, - default_kwargs: dict, - dump: bool, - out_file: Path, - force: bool, - ) -> List[str]: - serialized = [] - for schema_file in path.parent.glob("*.yaml"): - this_out = out_file.parent / (module_case(schema_file.stem) + ".py") - serialized.append( - self._build_unsplit(schema_file, versions, default_kwargs, dump, this_out, force) - ) - - # If there are dependent versions that also need to be built, do that now! - needed = [ - (module_case(ns), version_module_case(version)) - for ns, version in versions.items() - if version_module_case(version) not in self.available_versions.get(ns, []) - ] - for needs in needed: - needed_path = path.parents[2] / needs[0] / needs[1] / "namespace.yaml" - out_file_stem = out_file.parents[2] / needs[0] / needs[1] - for schema_file in needed_path.parent.glob("*.yaml"): - this_out = out_file_stem / (module_case(schema_file.stem) + ".py") - serialized.append( - self._build_unsplit( - schema_file, versions, default_kwargs, dump, this_out, force - ) - ) - - return serialized - - def _get_dependent_versions(self, path: Path) -> dict[str, str]: - """ - For a given namespace schema file, get the versions of any other schemas it imports - - Namespace imports will be in the importing schema like: - - imports: - -../../hdmf_common/v1_8_0/namespace - -../../hdmf_experimental/v0_5_0/namespace - - Returns: - dict[str,str]: A dictionary mapping a namespace to a version number - """ - schema = io.schema.load_yaml(path) - versions = {} - for i in schema["imports"]: - if i.startswith(".."): - import_path = (Path(path).parent / Path(i + ".yaml")).resolve() - imported_schema = io.schema.load_yaml(import_path) - versions[imported_schema["name"]] = imported_schema["version"] - return versions - - @classmethod - def module_name(self, namespace: str, version: str) -> str: - """Module name for the built module - - e.g.:: - - nwb_linkml.models.pydantic.{namespace}.{version} - """ - name_pieces = [ - "nwb_linkml", - "models", - "pydantic", - module_case(namespace), - version_module_case(version), - ] - module_name = ".".join(name_pieces) - return module_name - - def import_module(self, namespace: str, version: Optional[str] = None) -> ModuleType: - """ - Import a module within the temporary directory from its namespace and version - - In most cases, you're looking for :meth:`.PydanticProvider.get`, this method is - made available in case you don't want to accidentally build something - or invoke the rest of the provisioning system. - - Args: - namespace (str): Name of namespace - version (Optional[str]): Version to import, if None, - try and get the most recently built version. - - Returns: - :class:`types.ModuleType` - """ - # get latest version if None - if version is None: - version = self.available_versions[namespace][-1] - - path = self.namespace_path(namespace, version) / "namespace.py" - if not path.exists(): - raise ImportError(f"Module has not been built yet {path}") - module_name = self.module_name(namespace, version) - - # import module level first - when python does relative imports, - # it needs to have the parent modules imported separately - # this breaks split model creation when they are outside of the - # package repository (ie. loaded from an nwb file) because it tries - # to look for the containing namespace folder within the nwb_linkml package and fails - init_spec = importlib.util.spec_from_file_location(module_name, path.parent / "__init__.py") - init_module = importlib.util.module_from_spec(init_spec) - sys.modules[module_name] = init_module - init_spec.loader.exec_module(init_module) - - # then the namespace package - module_name = module_name + ".namespace" - spec = importlib.util.spec_from_file_location(module_name, path) - module = importlib.util.module_from_spec(spec) - sys.modules[module_name] = module - spec.loader.exec_module(module) - return module - - def get( - self, namespace: str, version: Optional[str] = None, allow_repo: Optional[bool] = None - ) -> ModuleType: - """ - Get the imported module for a given namespace and version. - - A given namespace will be stored in :data:`sys.modules` as - ``nwb_linkml.models.{namespace}``, - so first check if there is any already-imported module, and return that if so. - - Then we check in the temporary directory for an already-built ``namespace.py`` file - - Otherwise we pass arguments to :meth:`.PydanticProvider.build` and attempt to build them - before returning. - - Notes: - The imported modules shadow the "actual" - ``nwb_linkml.models`` module as would be imported from the usual location - within the package directory. - This is intentional, as models can then be used as if they were - integrated parts of the package, - and also so the active version of a namespace can be cleanly accessed - (ie. without ``from nwb_linkml.models.core import v2_2_0 as core`` ). - Accordingly, we assume that people will only be using a single version of NWB in a given - Python session. - - - Args: - namespace (str): Name of namespace to import. Must have either been previously built - with :meth:`.PydanticProvider.build` or - a matching namespace/version combo must be available to the - :class:`.LinkMLProvider` - version (Optional[str]): Version to import. If ``None``, - get the most recently build module - allow_repo (bool): Allow getting modules provided within - :mod:`nwb_linkml.models.pydantic` - - Returns: - The imported :class:`types.ModuleType` object that has all the built - classes at the root level. - - """ - if allow_repo is None: - allow_repo = self.allow_repo - - if version is None: - version = self.available_versions[namespace][-1] - - module_name = self.module_name(namespace, version) - namespace_name = module_name + ".namespace" - - if not allow_repo: - self._clear_package_imports() - - if namespace_name in sys.modules: - return sys.modules[namespace_name] - - try: - path = self.namespace_path(namespace, version, allow_repo) - except FileNotFoundError: - path = None - - if path is None or not path.exists(): - _ = self.build(namespace, version=version) - - module = self.import_module(namespace, version) - return module - - @staticmethod - def _clear_package_imports() -> None: - """ - When using allow_repo=False, delete any already-imported - namespaces from sys.modules that are within the nwb_linkml package - """ - # make sure we don't have any in-repo modules - repo_base = Path(__file__).parents[1] - deletes = [] - for k, v in sys.modules.items(): - if not k.startswith("nwb_linkml.models.pydantic"): - continue - try: - Path(v.__file__).relative_to(repo_base) - # if it is a subpath, delete it - deletes.append(k) - except ValueError: - # path is not a subpath - continue - for d in deletes: - del sys.modules[d] - - def get_class( - self, namespace: str, class_: str, version: Optional[str] = None - ) -> Type[BaseModel]: - """ - Get a class from a given namespace and version! - - Args: - namespace (str): Name of a namespace that has been previously built and cached, - otherwise we will attempt to build it from the - :data:`.providers.git.DEFAULT_REPOS` - class_ (str): Name of class to retrieve - version (Optional[str]): Optional version of the schema to retrieve from - - Returns: - :class:`pydantic.BaseModel` - """ - mod = self.get(namespace, version) - return getattr(mod, class_) - - -class EctopicModelFinder(MetaPathFinder): - """ - A meta path finder that allows the import machinery to find a model - package even if it might be outside the actual nwb_linkml namespace, - as occurs when building split models in a temporary directory. - - References: - - https://docs.python.org/3/reference/import.html#the-meta-path - - https://docs.python.org/3/library/importlib.html#importlib.abc.MetaPathFinder - - """ - - MODEL_STEM = "nwb_linkml.models.pydantic" - - def __init__(self, path: Path, *args, **kwargs): - super().__init__(*args, **kwargs) - self.path = path - - def find_spec( - self, fullname: str, path: Optional[str], target: Optional[ModuleType] = None - ) -> Optional[ModuleSpec]: - """If we are loading a generated pydantic module, return an importlib spec""" - if not fullname.startswith(self.MODEL_STEM): - return None - else: - # get submodule beneath model stem - submod = fullname.replace(self.MODEL_STEM, "").lstrip(".") - base_path = Path(self.path, *submod.split(".")) - - # switch if we're asked for a package or a module - mod_path = Path(str(base_path) + ".py") - pkg_path = base_path / "__init__.py" - if mod_path.exists(): - import_path = mod_path - elif pkg_path.exists(): - import_path = pkg_path - else: - return None - - spec = importlib.util.spec_from_file_location(fullname, import_path) - return spec +from nwb_linkml import adapters +from nwb_linkml.providers import LinkMLProvider, Provider, PydanticProvider class SchemaProvider(Provider): diff --git a/nwb_linkml/tests/test_providers/test_provider_schema.py b/nwb_linkml/tests/test_providers/test_provider_schema.py index f5c559d..de7c67a 100644 --- a/nwb_linkml/tests/test_providers/test_provider_schema.py +++ b/nwb_linkml/tests/test_providers/test_provider_schema.py @@ -8,8 +8,8 @@ from nptyping import Shape, UByte import nwb_linkml from nwb_linkml.maps.naming import version_module_case +from nwb_linkml.providers import LinkMLProvider, PydanticProvider from nwb_linkml.providers.git import DEFAULT_REPOS -from nwb_linkml.providers.schema import LinkMLProvider, PydanticProvider from nwb_linkml.types.ndarray import NDArray CORE_MODULES = ( diff --git a/scripts/generate_core.py b/scripts/generate_core.py index 3efbd76..efa9252 100644 --- a/scripts/generate_core.py +++ b/scripts/generate_core.py @@ -13,7 +13,7 @@ from rich.progress import Progress, SpinnerColumn, BarColumn, TextColumn, Column from rich import print from nwb_linkml.generators.pydantic import NWBPydanticGenerator -from nwb_linkml.providers.schema import LinkMLProvider, PydanticProvider +from nwb_linkml.providers import LinkMLProvider, PydanticProvider from nwb_linkml.providers.git import NWB_CORE_REPO, GitRepo from nwb_linkml.io import schema as io