Navigation
- index
- modules |
- next |
- previous |
- Python »
- 3.4.3 Documentation »
- The Python Standard Library »
- 31. Importing Modules »
New in version 3.1.
The purpose of the importlib package is two-fold. One is to provide the implementation of the import statement (and thus, by extension, the __import__() function) in Python source code. This provides an implementation of import which is portable to any Python interpreter. This also provides an implementation which is easier to comprehend than one implemented in a programming language other than Python.
Two, the components to implement import are exposed in this package, making it easier for users to create their own custom objects (known generically as an importer) to participate in the import process.
See also
An implementation of the built-in __import__() function.
Import a module. The name argument specifies what module to import in absolute or relative terms (e.g. either pkg.mod or ..mod). If the name is specified in relative terms, then the package argument must be set to the name of the package which is to act as the anchor for resolving the package name (e.g. import_module('..mod', 'pkg.subpkg') will import pkg.mod).
The import_module() function acts as a simplifying wrapper around importlib.__import__(). This means all semantics of the function are derived from importlib.__import__(), including requiring the package from which an import is occurring to have been previously imported (i.e., package must already be imported). The most important difference is that import_module() returns the specified package or module (e.g. pkg.mod), while __import__() returns the top-level package or module (e.g. pkg).
Changed in version 3.3: Parent packages are automatically imported.
Find the loader for a module, optionally within the specified path. If the module is in sys.modules, then sys.modules[name].__loader__ is returned (unless the loader would be None or is not set, in which case ValueError is raised). Otherwise a search using sys.meta_path is done. None is returned if no loader is found.
A dotted name does not have its parent’s implicitly imported as that requires loading them and that may not be desired. To properly import a submodule you will need to import all parent packages of the submodule and use the correct argument to path.
New in version 3.3.
Changed in version 3.4: If __loader__ is not set, raise ValueError, just like when the attribute is set to None.
Deprecated since version 3.4: Use importlib.util.find_spec() instead.
Invalidate the internal caches of finders stored at sys.meta_path. If a finder implements invalidate_caches() then it will be called to perform the invalidation. This function should be called if any modules are created/installed while your program is running to guarantee all finders will notice the new module’s existence.
New in version 3.3.
Reload a previously imported module. The argument must be a module object, so it must have been successfully imported before. This is useful if you have edited the module source file using an external editor and want to try out the new version without leaving the Python interpreter. The return value is the module object (which can be different if re-importing causes a different object to be placed in sys.modules).
When reload() is executed:
There are a number of other caveats:
When a module is reloaded, its dictionary (containing the module’s global variables) is retained. Redefinitions of names will override the old definitions, so this is generally not a problem. If the new version of a module does not define a name that was defined by the old version, the old definition remains. This feature can be used to the module’s advantage if it maintains a global table or cache of objects — with a try statement it can test for the table’s presence and skip its initialization if desired:
try:
cache
except NameError:
cache = {}
It is legal though generally not very useful to reload built-in or dynamically loaded modules (this is not true for e.g. sys, __main__, builtins and other key modules where reloading is frowned upon). In many cases, however, extension modules are not designed to be initialized more than once, and may fail in arbitrary ways when reloaded.
If a module imports objects from another module using from ... import ..., calling reload() for the other module does not redefine the objects imported from it — one way around this is to re-execute the from statement, another is to use import and qualified names (module.name) instead.
If a module instantiates instances of a class, reloading the module that defines the class does not affect the method definitions of the instances — they continue to use the old class definition. The same is true for derived classes.
New in version 3.4.
The importlib.abc module contains all of the core abstract base classes used by import. Some subclasses of the core abstract base classes are also provided to help in implementing the core ABCs.
ABC hierarchy:
object
+-- Finder (deprecated)
| +-- MetaPathFinder
| +-- PathEntryFinder
+-- Loader
+-- ResourceLoader --------+
+-- InspectLoader |
+-- ExecutionLoader --+
+-- FileLoader
+-- SourceLoader
An abstract base class representing a finder.
Deprecated since version 3.3: Use MetaPathFinder or PathEntryFinder instead.
An abstact method for finding a loader for the specified module. Originally specified in PEP 302, this method was meant for use in sys.meta_path and in the path-based import subsystem.
Changed in version 3.4: Returns None when called instead of raising NotImplementedError.
An abstract base class representing a meta path finder. For compatibility, this is a subclass of Finder.
New in version 3.3.
An abstract method for finding a spec for the specified module. If this is a top-level import, path will be None. Otherwise, this is a search for a subpackage or module and path will be the value of __path__ from the parent package. If a spec cannot be found, None is returned. When passed in, target is a module object that the finder may use to make a more educated about what spec to return.
New in version 3.4.
A legacy method for finding a loader for the specified module. If this is a top-level import, path will be None. Otherwise, this is a search for a subpackage or module and path will be the value of __path__ from the parent package. If a loader cannot be found, None is returned.
If find_spec() is defined, backwards-compatible functionality is provided.
Changed in version 3.4: Returns None when called instead of raising NotImplementedError. Can use find_spec() to provide functionality.
Deprecated since version 3.4: Use find_spec() instead.
An optional method which, when called, should invalidate any internal cache used by the finder. Used by importlib.invalidate_caches() when invalidating the caches of all finders on sys.meta_path.
Changed in version 3.4: Returns None when called instead of NotImplemented.
An abstract base class representing a path entry finder. Though it bears some similarities to MetaPathFinder, PathEntryFinder is meant for use only within the path-based import subsystem provided by PathFinder. This ABC is a subclass of Finder for compatibility reasons only.
New in version 3.3.
An abstract method for finding a spec for the specified module. The finder will search for the module only within the path entry to which it is assigned. If a spec cannot be found, None is returned. When passed in, target is a module object that the finder may use to make a more educated about what spec to return.
New in version 3.4.
A legacy method for finding a loader for the specified module. Returns a 2-tuple of (loader, portion) where portion is a sequence of file system locations contributing to part of a namespace package. The loader may be None while specifying portion to signify the contribution of the file system locations to a namespace package. An empty list can be used for portion to signify the loader is not part of a namespace package. If loader is None and portion is the empty list then no loader or location for a namespace package were found (i.e. failure to find anything for the module).
If find_spec() is defined then backwards-compatible functionality is provided.
Changed in version 3.4: Returns (None, []) instead of raising NotImplementedError. Uses find_spec() when available to provide functionality.
Deprecated since version 3.4: Use find_spec() instead.
A concrete implementation of Finder.find_module() which is equivalent to self.find_loader(fullname)[0].
Deprecated since version 3.4: Use find_spec() instead.
An optional method which, when called, should invalidate any internal cache used by the finder. Used by PathFinder.invalidate_caches() when invalidating the caches of all cached finders.
An abstract base class for a loader. See PEP 302 for the exact definition for a loader.
An optional method that returns the module object to use when importing a module. create_module() may also return None, indicating that the default module creation should take place instead.
New in version 3.4.
An abstract method that executes the module in its own namespace when a module is imported or reloaded. The module should already be initialized when exec_module() is called.
New in version 3.4.
A legacy method for loading a module. If the module cannot be loaded, ImportError is raised, otherwise the loaded module is returned.
If the requested module already exists in sys.modules, that module should be used and reloaded. Otherwise the loader should create a new module and insert it into sys.modules before any loading begins, to prevent recursion from the import. If the loader inserted a module and the load fails, it must be removed by the loader from sys.modules; modules already in sys.modules before the loader began execution should be left alone (see importlib.util.module_for_loader()).
The loader should set several attributes on the module. (Note that some of these attributes can change when a module is reloaded):
The name of the module.
The path to where the module data is stored (not set for built-in modules).
The path to where a compiled version of the module is/should be stored (not set when the attribute would be inappropriate).
A list of strings specifying the search path within a package. This attribute is not set on modules.
The parent package for the module/package. If the module is top-level then it has a value of the empty string. The importlib.util.module_for_loader() decorator can handle the details for __package__.
The loader used to load the module. The importlib.util.module_for_loader() decorator can handle the details for __package__.
When exec_module() is available then backwards-compatible functionality is provided.
Changed in version 3.4: Raise ImportError when called instead of NotImplementedError. Functionality provided when exec_module() is available.
Deprecated since version 3.4: The recommended API for loading a module is exec_module() (and optionally create_module()). Loaders should implement it instead of load_module(). The import machinery takes care of all the other responsibilities of load_module() when exec_module() is implemented.
A legacy method which when implemented calculates and returns the given module’s repr, as a string. The module type’s default repr() will use the result of this method as appropriate.
New in version 3.3.
Changed in version 3.4: Made optional instead of an abstractmethod.
Deprecated since version 3.4: The import machinery now takes care of this automatically.
An abstract base class for a loader which implements the optional PEP 302 protocol for loading arbitrary resources from the storage back-end.
An abstract method to return the bytes for the data located at path. Loaders that have a file-like storage back-end that allows storing arbitrary data can implement this abstract method to give direct access to the data stored. OSError is to be raised if the path cannot be found. The path is expected to be constructed using a module’s __file__ attribute or an item from a package’s __path__.
Changed in version 3.4: Raises OSError instead of NotImplementedError.
An abstract base class for a loader which implements the optional PEP 302 protocol for loaders that inspect modules.
Return the code object for a module, or None if the module does not have a code object (as would be the case, for example, for a built-in module). Raise an ImportError if loader cannot find the requested module.
Note
While the method has a default implementation, it is suggested that it be overridden if possible for performance.
Changed in version 3.4: No longer abstract and a concrete implementation is provided.
An abstract method to return the source of a module. It is returned as a text string using universal newlines, translating all recognized line separators into '\n' characters. Returns None if no source is available (e.g. a built-in module). Raises ImportError if the loader cannot find the module specified.
Changed in version 3.4: Raises ImportError instead of NotImplementedError.
An abstract method to return a true value if the module is a package, a false value otherwise. ImportError is raised if the loader cannot find the module.
Changed in version 3.4: Raises ImportError instead of NotImplementedError.
Create a code object from Python source.
The data argument can be whatever the compile() function supports (i.e. string or bytes). The path argument should be the “path” to where the source code originated from, which can be an abstract concept (e.g. location in a zip file).
New in version 3.4.
Implementation of Loader.exec_module().
New in version 3.4.
Implementation of Loader.load_module().
Deprecated since version 3.4: use exec_module() instead.
An abstract base class which inherits from InspectLoader that, when implemented, helps a module to be executed as a script. The ABC represents an optional PEP 302 protocol.
An abstract method that is to return the value of __file__ for the specified module. If no path is available, ImportError is raised.
If source code is available, then the method should return the path to the source file, regardless of whether a bytecode was used to load the module.
Changed in version 3.4: Raises ImportError instead of NotImplementedError.
An abstract base class which inherits from ResourceLoader and ExecutionLoader, providing concrete implementations of ResourceLoader.get_data() and ExecutionLoader.get_filename().
The fullname argument is a fully resolved name of the module the loader is to handle. The path argument is the path to the file for the module.
New in version 3.3.
The name of the module the loader can handle.
Path to the file of the module.
Calls super’s load_module().
Deprecated since version 3.4: Use Loader.exec_module() instead.
Reads path as a binary file and returns the bytes from it.
An abstract base class for implementing source (and optionally bytecode) file loading. The class inherits from both ResourceLoader and ExecutionLoader, requiring the implementation of:
Should only return the path to the source file; sourceless loading is not supported.
The abstract methods defined by this class are to add optional bytecode file support. Not implementing these optional methods (or causing them to raise NotImplementedError) causes the loader to only work with source code. Implementing the methods allows the loader to work with source and bytecode files; it does not allow for sourceless loading where only bytecode is provided. Bytecode files are an optimization to speed up loading by removing the parsing step of Python’s compiler, and so no bytecode-specific API is exposed.
Optional abstract method which returns a dict containing metadata about the specifed path. Supported dictionary keys are:
Any other keys in the dictionary are ignored, to allow for future extensions. If the path cannot be handled, OSError is raised.
New in version 3.3.
Changed in version 3.4: Raise OSError instead of NotImplementedError.
Optional abstract method which returns the modification time for the specified path.
Deprecated since version 3.3: This method is deprecated in favour of path_stats(). You don’t have to implement it, but it is still available for compatibility purposes. Raise OSError if the path cannot be handled.
Changed in version 3.4: Raise OSError instead of NotImplementedError.
Optional abstract method which writes the specified bytes to a file path. Any intermediate directories which do not exist are to be created automatically.
When writing to the path fails because the path is read-only (errno.EACCES/PermissionError), do not propagate the exception.
Changed in version 3.4: No longer raises NotImplementedError when called.
Concrete implementation of InspectLoader.get_code().
Concrete implementation of Loader.exec_module().
New in version 3.4.
Concrete implementation of Loader.load_module().
Deprecated since version 3.4: Use exec_module() instead.
Concrete implementation of InspectLoader.get_source().
Concrete implementation of InspectLoader.is_package(). A module is determined to be a package if its file path (as provided by ExecutionLoader.get_filename()) is a file named __init__ when the file extension is removed and the module name itself does not end in __init__.
This module contains the various objects that help import find and load modules.
A list of strings representing the recognized file suffixes for source modules.
New in version 3.3.
A list of strings representing the file suffixes for non-optimized bytecode modules.
New in version 3.3.
A list of strings representing the file suffixes for optimized bytecode modules.
New in version 3.3.
A list of strings representing the recognized file suffixes for bytecode modules. Set to either DEBUG_BYTECODE_SUFFIXES or OPTIMIZED_BYTECODE_SUFFIXES based on whether __debug__ is true.
New in version 3.3.
A list of strings representing the recognized file suffixes for extension modules.
New in version 3.3.
Returns a combined list of strings representing all file suffixes for modules recognized by the standard import machinery. This is a helper for code which simply needs to know if a filesystem path potentially refers to a module without needing any details on the kind of module (for example, inspect.getmodulename())
New in version 3.3.
An importer for built-in modules. All known built-in modules are listed in sys.builtin_module_names. This class implements the importlib.abc.MetaPathFinder and importlib.abc.InspectLoader ABCs.
Only class methods are defined by this class to alleviate the need for instantiation.
Note
Due to limitations in the extension module C-API, for now BuiltinImporter does not implement Loader.exec_module().
An importer for frozen modules. This class implements the importlib.abc.MetaPathFinder and importlib.abc.InspectLoader ABCs.
Only class methods are defined by this class to alleviate the need for instantiation.
Finder for modules declared in the Windows registry. This class implements the importlib.abc.Finder ABC.
Only class methods are defined by this class to alleviate the need for instantiation.
New in version 3.3.
A Finder for sys.path and package __path__ attributes. This class implements the importlib.abc.MetaPathFinder ABC.
Only class methods are defined by this class to alleviate the need for instantiation.
Class method that attempts to find a spec for the module specified by fullname on sys.path or, if defined, on path. For each path entry that is searched, sys.path_importer_cache is checked. If a non-false object is found then it is used as the path entry finder to look for the module being searched for. If no entry is found in sys.path_importer_cache, then sys.path_hooks is searched for a finder for the path entry and, if found, is stored in sys.path_importer_cache along with being queried about the module. If no finder is ever found then None is both stored in the cache and returned.
New in version 3.4.
A legacy wrapper around find_spec().
Deprecated since version 3.4: Use find_spec() instead.
Calls importlib.abc.PathEntryFinder.invalidate_caches() on all finders stored in sys.path_importer_cache.
Changed in version 3.4: Calls objects in sys.path_hooks with the current working directory for '' (i.e. the empty string).
A concrete implementation of importlib.abc.PathEntryFinder which caches results from the file system.
The path argument is the directory for which the finder is in charge of searching.
The loader_details argument is a variable number of 2-item tuples each containing a loader and a sequence of file suffixes the loader recognizes. The loaders are expected to be callables which accept two arguments of the module’s name and the path to the file found.
The finder will cache the directory contents as necessary, making stat calls for each module search to verify the cache is not outdated. Because cache staleness relies upon the granularity of the operating system’s state information of the file system, there is a potential race condition of searching for a module, creating a new file, and then searching for the module the new file represents. If the operations happen fast enough to fit within the granularity of stat calls, then the module search will fail. To prevent this from happening, when you create a module dynamically, make sure to call importlib.invalidate_caches().
New in version 3.3.
The path the finder will search in.
Attempt to find the spec to handle fullname within path.
New in version 3.4.
Clear out the internal cache.
A class method which returns a closure for use on sys.path_hooks. An instance of FileFinder is returned by the closure using the path argument given to the closure directly and loader_details indirectly.
If the argument to the closure is not an existing directory, ImportError is raised.
A concrete implementation of importlib.abc.SourceLoader by subclassing importlib.abc.FileLoader and providing some concrete implementations of other methods.
New in version 3.3.
The name of the module that this loader will handle.
The path to the source file.
Concrete implementation of importlib.abc.SourceLoader.path_stats().
Concrete implementation of importlib.abc.SourceLoader.set_data().
Concrete implementation of importlib.abc.Loader.load_module() where specifying the name of the module to load is optional.
A concrete implementation of importlib.abc.FileLoader which can import bytecode files (i.e. no source code files exist).
Please note that direct use of bytecode files (and thus not source code files) inhibits your modules from being usable by all Python implementations or new versions of Python which change the bytecode format.
New in version 3.3.
The name of the module the loader will handle.
The path to the bytecode file.
Returns None as bytecode files have no source when this loader is used.
Concrete implementation of importlib.abc.Loader.load_module() where specifying the name of the module to load is optional.
A concrete implementation of importlib.abc.ExecutionLoader for extension modules.
The fullname argument specifies the name of the module the loader is to support. The path argument is the path to the extension module’s file.
New in version 3.3.
Name of the module the loader supports.
Path to the extension module.
Loads the extension module if and only if fullname is the same as name or is None.
Note
Due to limitations in the extension module C-API, for now ExtensionFileLoader does not implement Loader.exec_module().
Returns True if the file path points to a package’s __init__ module based on EXTENSION_SUFFIXES.
Returns None as extension modules lack a code object.
Returns None as extension modules do not have source code.
A specification for a module’s import-system-related state.
New in version 3.4.
(__name__)
A string for the fully-qualified name of the module.
(__loader__)
The loader to use for loading. For namespace packages this should be set to None.
(__file__)
Name of the place from which the module is loaded, e.g. “builtin” for built-in modules and the filename for modules loaded from source. Normally “origin” should be set, but it may be None (the default) which indicates it is unspecified.
(__path__)
List of strings for where to find submodules, if a package (None otherwise).
Container of extra module-specific data for use during loading (or None).
(__cached__)
String for where the compiled module should be stored (or None).
(__package__)
(Read-only) Fully-qualified name of the package to which the module belongs as a submodule (or None).
Boolean indicating whether or not the module’s “origin” attribute refers to a loadable location.
This module contains the various objects that help in the construction of an importer.
The bytes which represent the bytecode version number. If you need help with loading/writing bytecode then consider importlib.abc.SourceLoader.
New in version 3.4.
Return the PEP 3147 path to the byte-compiled file associated with the source path. For example, if path is /foo/bar/baz.py the return value would be /foo/bar/__pycache__/baz.cpython-32.pyc for Python 3.2. The cpython-32 string comes from the current magic tag (see get_tag(); if sys.implementation.cache_tag is not defined then NotImplementedError will be raised). The returned path will end in .pyc when __debug__ is True or .pyo for an optimized Python (i.e. __debug__ is False). By passing in True or False for debug_override you can override the system’s value for __debug__ for extension selection.
path need not exist.
New in version 3.4.
Given the path to a PEP 3147 file name, return the associated source code file path. For example, if path is /foo/bar/__pycache__/baz.cpython-32.pyc the returned path would be /foo/bar/baz.py. path need not exist, however if it does not conform to PEP 3147 format, a ValueError is raised. If sys.implementation.cache_tag is not defined, NotImplementedError is raised.
New in version 3.4.
Decode the given bytes representing source code and return it as a string with universal newlines (as required by importlib.abc.InspectLoader.get_source()).
New in version 3.4.
Resolve a relative module name to an absolute one.
If name has no leading dots, then name is simply returned. This allows for usage such as importlib.util.resolve_name('sys', __package__) without doing a check to see if the package argument is needed.
ValueError is raised if name is a relative module name but package is a false value (e.g. None or the empty string). ValueError is also raised a relative name would escape its containing package (e.g. requesting ..bacon from within the spam package).
New in version 3.3.
Find the spec for a module, optionally relative to the specified package name. If the module is in sys.modules, then sys.modules[name].__spec__ is returned (unless the spec would be None or is not set, in which case ValueError is raised). Otherwise a search using sys.meta_path is done. None is returned if no spec is found.
If name is for a submodule (contains a dot), the parent module is automatically imported.
name and package work the same as for import_module().
New in version 3.4.
A decorator for importlib.abc.Loader.load_module() to handle selecting the proper module object to load with. The decorated method is expected to have a call signature taking two positional arguments (e.g. load_module(self, module)) for which the second argument will be the module object to be used by the loader. Note that the decorator will not work on static methods because of the assumption of two arguments.
The decorated method will take in the name of the module to be loaded as expected for a loader. If the module is not found in sys.modules then a new one is constructed. Regardless of where the module came from, __loader__ set to self and __package__ is set based on what importlib.abc.InspectLoader.is_package() returns (if available). These attributes are set unconditionally to support reloading.
If an exception is raised by the decorated method and a module was added to sys.modules, then the module will be removed to prevent a partially initialized module from being in left in sys.modules. If the module was already in sys.modules then it is left alone.
Changed in version 3.3: __loader__ and __package__ are automatically set (when possible).
Changed in version 3.4: Set __name__, __loader__ __package__ unconditionally to support reloading.
Deprecated since version 3.4: The import machinery now directly performs all the functionality provided by this function.
A decorator for importlib.abc.Loader.load_module() to set the __loader__ attribute on the returned module. If the attribute is already set the decorator does nothing. It is assumed that the first positional argument to the wrapped method (i.e. self) is what __loader__ should be set to.
Changed in version 3.4: Set __loader__ if set to None, as if the attribute does not exist.
Deprecated since version 3.4: The import machinery takes care of this automatically.
A decorator for importlib.abc.Loader.load_module() to set the __package__ attribute on the returned module. If __package__ is set and has a value other than None it will not be changed.
Deprecated since version 3.4: The import machinery takes care of this automatically.
A factory function for creating a ModuleSpec instance based on a loader. The parameters have the same meaning as they do for ModuleSpec. The function uses available loader APIs, such as InspectLoader.is_package(), to fill in any missing information on the spec.
New in version 3.4.
A factory function for creating a ModuleSpec instance based on the path to a file. Missing information will be filled in on the spec by making use of loader APIs and by the implication that the module will be file-based.
New in version 3.4.