bpo-42129: Add support for resources in namespaces (GH-24670)
* Unify behavior in ResourceReaderDefaultsTests and align with the behavior found in importlib_resources.
* Equip NamespaceLoader with a NamespaceReader.
* Apply changes from importlib_resources 5.0.4
diff --git a/Lib/importlib/_adapters.py b/Lib/importlib/_adapters.py
new file mode 100644
index 0000000..eedde49
--- /dev/null
+++ b/Lib/importlib/_adapters.py
@@ -0,0 +1,82 @@
+from contextlib import suppress
+
+from . import abc
+
+
+class SpecLoaderAdapter:
+ """
+ Adapt a package spec to adapt the underlying loader.
+ """
+
+ def __init__(self, spec, adapter=lambda spec: spec.loader):
+ self.spec = spec
+ self.loader = adapter(spec)
+
+ def __getattr__(self, name):
+ return getattr(self.spec, name)
+
+
+class TraversableResourcesLoader:
+ """
+ Adapt a loader to provide TraversableResources.
+ """
+
+ def __init__(self, spec):
+ self.spec = spec
+
+ def get_resource_reader(self, name):
+ return DegenerateFiles(self.spec)._native()
+
+
+class DegenerateFiles:
+ """
+ Adapter for an existing or non-existant resource reader
+ to provide a degenerate .files().
+ """
+
+ class Path(abc.Traversable):
+ def iterdir(self):
+ return iter(())
+
+ def is_dir(self):
+ return False
+
+ is_file = exists = is_dir # type: ignore
+
+ def joinpath(self, other):
+ return DegenerateFiles.Path()
+
+ def name(self):
+ return ''
+
+ def open(self):
+ raise ValueError()
+
+ def __init__(self, spec):
+ self.spec = spec
+
+ @property
+ def _reader(self):
+ with suppress(AttributeError):
+ return self.spec.loader.get_resource_reader(self.spec.name)
+
+ def _native(self):
+ """
+ Return the native reader if it supports files().
+ """
+ reader = self._reader
+ return reader if hasattr(reader, 'files') else self
+
+ def __getattr__(self, attr):
+ return getattr(self._reader, attr)
+
+ def files(self):
+ return DegenerateFiles.Path()
+
+
+def wrap_spec(package):
+ """
+ Construct a package spec with traversable compatibility
+ on the spec/loader/reader.
+ """
+ return SpecLoaderAdapter(package.__spec__, TraversableResourcesLoader)
diff --git a/Lib/importlib/_bootstrap_external.py b/Lib/importlib/_bootstrap_external.py
index 3396d5d..2dab45d 100644
--- a/Lib/importlib/_bootstrap_external.py
+++ b/Lib/importlib/_bootstrap_external.py
@@ -1259,6 +1259,10 @@ def load_module(self, fullname):
# Warning implemented in _load_module_shim().
return _bootstrap._load_module_shim(self, fullname)
+ def get_resource_reader(self, module):
+ from importlib.readers import NamespaceReader
+ return NamespaceReader(self._path)
+
# Finders #####################################################################
diff --git a/Lib/importlib/_common.py b/Lib/importlib/_common.py
index 71ce6af..0338304 100644
--- a/Lib/importlib/_common.py
+++ b/Lib/importlib/_common.py
@@ -9,6 +9,8 @@
from typing import Union, Any, Optional
from .abc import ResourceReader
+from ._adapters import wrap_spec
+
Package = Union[types.ModuleType, str]
@@ -43,18 +45,15 @@ def get_resource_reader(package):
# zipimport.zipimporter does not support weak references, resulting in a
# TypeError. That seems terrible.
spec = package.__spec__
- reader = getattr(spec.loader, 'get_resource_reader', None)
+ reader = getattr(spec.loader, 'get_resource_reader', None) # type: ignore
if reader is None:
return None
- return reader(spec.name)
+ return reader(spec.name) # type: ignore
def resolve(cand):
# type: (Package) -> types.ModuleType
- return (
- cand if isinstance(cand, types.ModuleType)
- else importlib.import_module(cand)
- )
+ return cand if isinstance(cand, types.ModuleType) else importlib.import_module(cand)
def get_package(package):
@@ -64,7 +63,7 @@ def get_package(package):
Raise an exception if the resolved module is not a package.
"""
resolved = resolve(package)
- if resolved.__spec__.submodule_search_locations is None:
+ if wrap_spec(resolved).submodule_search_locations is None:
raise TypeError('{!r} is not a package'.format(package))
return resolved
@@ -74,7 +73,7 @@ def from_package(package):
Return a Traversable object for the given package.
"""
- spec = package.__spec__
+ spec = wrap_spec(package)
reader = spec.loader.get_resource_reader(spec.name)
return reader.files()
diff --git a/Lib/importlib/abc.py b/Lib/importlib/abc.py
index 55e7088..4be51e2 100644
--- a/Lib/importlib/abc.py
+++ b/Lib/importlib/abc.py
@@ -14,6 +14,7 @@
from ._abc import Loader
import abc
import warnings
+from typing import BinaryIO, Iterable, Text
from typing import Protocol, runtime_checkable
@@ -297,49 +298,45 @@ def set_data(self, path, data):
class ResourceReader(metaclass=abc.ABCMeta):
-
- """Abstract base class to provide resource-reading support.
-
- Loaders that support resource reading are expected to implement
- the ``get_resource_reader(fullname)`` method and have it either return None
- or an object compatible with this ABC.
- """
+ """Abstract base class for loaders to provide resource reading support."""
@abc.abstractmethod
- def open_resource(self, resource):
+ def open_resource(self, resource: Text) -> BinaryIO:
"""Return an opened, file-like object for binary reading.
- The 'resource' argument is expected to represent only a file name
- and thus not contain any subdirectory components.
-
+ The 'resource' argument is expected to represent only a file name.
If the resource cannot be found, FileNotFoundError is raised.
"""
+ # This deliberately raises FileNotFoundError instead of
+ # NotImplementedError so that if this method is accidentally called,
+ # it'll still do the right thing.
raise FileNotFoundError
@abc.abstractmethod
- def resource_path(self, resource):
+ def resource_path(self, resource: Text) -> Text:
"""Return the file system path to the specified resource.
- The 'resource' argument is expected to represent only a file name
- and thus not contain any subdirectory components.
-
+ The 'resource' argument is expected to represent only a file name.
If the resource does not exist on the file system, raise
FileNotFoundError.
"""
+ # This deliberately raises FileNotFoundError instead of
+ # NotImplementedError so that if this method is accidentally called,
+ # it'll still do the right thing.
raise FileNotFoundError
@abc.abstractmethod
- def is_resource(self, name):
- """Return True if the named 'name' is consider a resource."""
+ def is_resource(self, path: Text) -> bool:
+ """Return True if the named 'path' is a resource.
+
+ Files are resources, directories are not.
+ """
raise FileNotFoundError
@abc.abstractmethod
- def contents(self):
- """Return an iterable of strings over the contents of the package."""
- return []
-
-
-_register(ResourceReader, machinery.SourceFileLoader)
+ def contents(self) -> Iterable[str]:
+ """Return an iterable of entries in `package`."""
+ raise FileNotFoundError
@runtime_checkable
@@ -355,26 +352,28 @@ def iterdir(self):
Yield Traversable objects in self
"""
- @abc.abstractmethod
def read_bytes(self):
"""
Read contents of self as bytes
"""
+ with self.open('rb') as strm:
+ return strm.read()
- @abc.abstractmethod
def read_text(self, encoding=None):
"""
- Read contents of self as bytes
+ Read contents of self as text
"""
+ with self.open(encoding=encoding) as strm:
+ return strm.read()
@abc.abstractmethod
- def is_dir(self):
+ def is_dir(self) -> bool:
"""
Return True if self is a dir
"""
@abc.abstractmethod
- def is_file(self):
+ def is_file(self) -> bool:
"""
Return True if self is a file
"""
@@ -385,11 +384,11 @@ def joinpath(self, child):
Return Traversable child in self
"""
- @abc.abstractmethod
def __truediv__(self, child):
"""
Return Traversable child in self
"""
+ return self.joinpath(child)
@abc.abstractmethod
def open(self, mode='r', *args, **kwargs):
@@ -402,14 +401,18 @@ def open(self, mode='r', *args, **kwargs):
"""
@abc.abstractproperty
- def name(self):
- # type: () -> str
+ def name(self) -> str:
"""
The base name of this object without any parent references.
"""
class TraversableResources(ResourceReader):
+ """
+ The required interface for providing traversable
+ resources.
+ """
+
@abc.abstractmethod
def files(self):
"""Return a Traversable object for the loaded package."""
diff --git a/Lib/importlib/readers.py b/Lib/importlib/readers.py
index 74a63e4..535c828 100644
--- a/Lib/importlib/readers.py
+++ b/Lib/importlib/readers.py
@@ -1,8 +1,13 @@
+import collections
import zipfile
import pathlib
from . import abc
+def remove_duplicates(items):
+ return iter(collections.OrderedDict.fromkeys(items))
+
+
class FileReader(abc.TraversableResources):
def __init__(self, loader):
self.path = pathlib.Path(loader.path).parent
@@ -39,3 +44,80 @@ def is_resource(self, path):
def files(self):
return zipfile.Path(self.archive, self.prefix)
+
+
+class MultiplexedPath(abc.Traversable):
+ """
+ Given a series of Traversable objects, implement a merged
+ version of the interface across all objects. Useful for
+ namespace packages which may be multihomed at a single
+ name.
+ """
+
+ def __init__(self, *paths):
+ self._paths = list(map(pathlib.Path, remove_duplicates(paths)))
+ if not self._paths:
+ message = 'MultiplexedPath must contain at least one path'
+ raise FileNotFoundError(message)
+ if not all(path.is_dir() for path in self._paths):
+ raise NotADirectoryError('MultiplexedPath only supports directories')
+
+ def iterdir(self):
+ visited = []
+ for path in self._paths:
+ for file in path.iterdir():
+ if file.name in visited:
+ continue
+ visited.append(file.name)
+ yield file
+
+ def read_bytes(self):
+ raise FileNotFoundError(f'{self} is not a file')
+
+ def read_text(self, *args, **kwargs):
+ raise FileNotFoundError(f'{self} is not a file')
+
+ def is_dir(self):
+ return True
+
+ def is_file(self):
+ return False
+
+ def joinpath(self, child):
+ # first try to find child in current paths
+ for file in self.iterdir():
+ if file.name == child:
+ return file
+ # if it does not exist, construct it with the first path
+ return self._paths[0] / child
+
+ __truediv__ = joinpath
+
+ def open(self, *args, **kwargs):
+ raise FileNotFoundError('{} is not a file'.format(self))
+
+ def name(self):
+ return self._paths[0].name
+
+ def __repr__(self):
+ return 'MultiplexedPath({})'.format(
+ ', '.join("'{}'".format(path) for path in self._paths)
+ )
+
+
+class NamespaceReader(abc.TraversableResources):
+ def __init__(self, namespace_path):
+ if 'NamespacePath' not in str(namespace_path):
+ raise ValueError('Invalid path')
+ self.path = MultiplexedPath(*list(namespace_path))
+
+ def resource_path(self, resource):
+ """
+ Return the file system path to prevent
+ `resources.path()` from creating a temporary
+ copy.
+ """
+ return str(self.path.joinpath(resource))
+
+ def files(self):
+ return self.path
diff --git a/Lib/importlib/resources.py b/Lib/importlib/resources.py
index 4169171..db0e0c0 100644
--- a/Lib/importlib/resources.py
+++ b/Lib/importlib/resources.py
@@ -3,8 +3,10 @@
from . import _common
from ._common import as_file, files
+from .abc import ResourceReader
from contextlib import suppress
from importlib.abc import ResourceLoader
+from importlib.machinery import ModuleSpec
from io import BytesIO, TextIOWrapper
from pathlib import Path
from types import ModuleType
@@ -18,6 +20,7 @@
__all__ = [
'Package',
'Resource',
+ 'ResourceReader',
'as_file',
'contents',
'files',
@@ -27,7 +30,7 @@
'path',
'read_binary',
'read_text',
- ]
+]
Package = Union[str, ModuleType]
@@ -41,36 +44,45 @@ def open_binary(package: Package, resource: Resource) -> BinaryIO:
reader = _common.get_resource_reader(package)
if reader is not None:
return reader.open_resource(resource)
- absolute_package_path = os.path.abspath(
- package.__spec__.origin or 'non-existent file')
- package_path = os.path.dirname(absolute_package_path)
- full_path = os.path.join(package_path, resource)
- try:
- return open(full_path, mode='rb')
- except OSError:
- # Just assume the loader is a resource loader; all the relevant
- # importlib.machinery loaders are and an AttributeError for
- # get_data() will make it clear what is needed from the loader.
- loader = cast(ResourceLoader, package.__spec__.loader)
- data = None
- if hasattr(package.__spec__.loader, 'get_data'):
- with suppress(OSError):
- data = loader.get_data(full_path)
- if data is None:
- package_name = package.__spec__.name
- message = '{!r} resource not found in {!r}'.format(
- resource, package_name)
- raise FileNotFoundError(message)
- return BytesIO(data)
+ spec = cast(ModuleSpec, package.__spec__)
+ # Using pathlib doesn't work well here due to the lack of 'strict'
+ # argument for pathlib.Path.resolve() prior to Python 3.6.
+ if spec.submodule_search_locations is not None:
+ paths = spec.submodule_search_locations
+ elif spec.origin is not None:
+ paths = [os.path.dirname(os.path.abspath(spec.origin))]
+
+ for package_path in paths:
+ full_path = os.path.join(package_path, resource)
+ try:
+ return open(full_path, mode='rb')
+ except OSError:
+ # Just assume the loader is a resource loader; all the relevant
+ # importlib.machinery loaders are and an AttributeError for
+ # get_data() will make it clear what is needed from the loader.
+ loader = cast(ResourceLoader, spec.loader)
+ data = None
+ if hasattr(spec.loader, 'get_data'):
+ with suppress(OSError):
+ data = loader.get_data(full_path)
+ if data is not None:
+ return BytesIO(data)
+
+ raise FileNotFoundError(
+ '{!r} resource not found in {!r}'.format(resource, spec.name)
+ )
-def open_text(package: Package,
- resource: Resource,
- encoding: str = 'utf-8',
- errors: str = 'strict') -> TextIO:
+def open_text(
+ package: Package,
+ resource: Resource,
+ encoding: str = 'utf-8',
+ errors: str = 'strict',
+) -> TextIO:
"""Return a file-like object opened for text reading of the resource."""
return TextIOWrapper(
- open_binary(package, resource), encoding=encoding, errors=errors)
+ open_binary(package, resource), encoding=encoding, errors=errors
+ )
def read_binary(package: Package, resource: Resource) -> bytes:
@@ -79,10 +91,12 @@ def read_binary(package: Package, resource: Resource) -> bytes:
return fp.read()
-def read_text(package: Package,
- resource: Resource,
- encoding: str = 'utf-8',
- errors: str = 'strict') -> str:
+def read_text(
+ package: Package,
+ resource: Resource,
+ encoding: str = 'utf-8',
+ errors: str = 'strict',
+) -> str:
"""Return the decoded string of the resource.
The decoding-related arguments have the same semantics as those of
@@ -93,8 +107,9 @@ def read_text(package: Package,
def path(
- package: Package, resource: Resource,
- ) -> 'ContextManager[Path]':
+ package: Package,
+ resource: Resource,
+) -> 'ContextManager[Path]':
"""A context manager providing a file path object to the resource.
If the resource does not already exist on its own on the file system,
@@ -106,15 +121,17 @@ def path(
reader = _common.get_resource_reader(_common.get_package(package))
return (
_path_from_reader(reader, _common.normalize_path(resource))
- if reader else
- _common.as_file(
- _common.files(package).joinpath(_common.normalize_path(resource)))
+ if reader
+ else _common.as_file(
+ _common.files(package).joinpath(_common.normalize_path(resource))
)
+ )
def _path_from_reader(reader, resource):
- return _path_from_resource_path(reader, resource) or \
- _path_from_open_resource(reader, resource)
+ return _path_from_resource_path(reader, resource) or _path_from_open_resource(
+ reader, resource
+ )
def _path_from_resource_path(reader, resource):
@@ -154,15 +171,10 @@ def contents(package: Package) -> Iterable[str]:
reader = _common.get_resource_reader(package)
if reader is not None:
return _ensure_sequence(reader.contents())
- # Is the package a namespace package? By definition, namespace packages
- # cannot have resources.
- namespace = (
- package.__spec__.origin is None or
- package.__spec__.origin == 'namespace'
- )
- if namespace or not package.__spec__.has_location:
- return ()
- return list(item.name for item in _common.from_package(package).iterdir())
+ transversable = _common.from_package(package)
+ if transversable.is_dir():
+ return list(item.name for item in transversable.iterdir())
+ return []
@singledispatch