cpython/Lib/importlib/_bootstrap.py

1461 lines
50 KiB
Python
Raw Normal View History

"""Core implementation of import.
This module is NOT meant to be directly imported! It has been designed such
that it can be bootstrapped into Python as the implementation of import. As
such it requires the injection of specific modules and attributes in order to
work. One should use importlib as the public-facing version of this module.
"""
#
# IMPORTANT: Whenever making changes to this module, be sure to run
# a top-level make in order to get the frozen version of the module
# update. Not doing so, will result in the Makefile to fail for
# all others who don't have a ./python around to freeze the module
# in the early stages of compilation.
#
# See importlib._setup() for what is injected into the global namespace.
# When editing this code be aware that code executed at import time CANNOT
# reference any injected objects! This includes not only global code but also
# anything specified at the class level.
# XXX Make sure all public names have no single leading underscore and all
# others do.
# Bootstrap-related code ######################################################
_CASE_INSENSITIVE_PLATFORMS = 'win', 'cygwin', 'darwin'
2012-02-20 08:52:47 -04:00
def _make_relax_case():
if sys.platform.startswith(_CASE_INSENSITIVE_PLATFORMS):
def _relax_case():
2012-02-20 16:06:35 -04:00
"""True if filenames must be checked case-insensitively."""
return b'PYTHONCASEOK' in _os.environ
else:
def _relax_case():
2012-02-20 16:06:35 -04:00
"""True if filenames must be checked case-insensitively."""
return False
return _relax_case
# TODO: Expose from marshal
def _w_long(x):
"""Convert a 32-bit integer to little-endian.
XXX Temporary until marshal's long functions are exposed.
"""
x = int(x)
int_bytes = []
int_bytes.append(x & 0xFF)
int_bytes.append((x >> 8) & 0xFF)
int_bytes.append((x >> 16) & 0xFF)
int_bytes.append((x >> 24) & 0xFF)
return bytearray(int_bytes)
# TODO: Expose from marshal
def _r_long(int_bytes):
"""Convert 4 bytes in little-endian to an integer.
XXX Temporary until marshal's long function are exposed.
"""
x = int_bytes[0]
x |= int_bytes[1] << 8
x |= int_bytes[2] << 16
x |= int_bytes[3] << 24
return x
def _path_join(*path_parts):
"""Replacement for os.path.join()."""
new_parts = []
for part in path_parts:
if not part:
continue
new_parts.append(part)
if part[-1] not in path_separators:
new_parts.append(path_sep)
return ''.join(new_parts[:-1]) # Drop superfluous path separator.
def _path_split(path):
"""Replacement for os.path.split()."""
for x in reversed(path):
if x in path_separators:
sep = x
break
else:
sep = path_sep
front, _, tail = path.rpartition(sep)
return front, tail
def _path_is_mode_type(path, mode):
"""Test whether the path is the specified mode type."""
try:
stat_info = _os.stat(path)
except OSError:
return False
return (stat_info.st_mode & 0o170000) == mode
# XXX Could also expose Modules/getpath.c:isfile()
def _path_isfile(path):
"""Replacement for os.path.isfile."""
return _path_is_mode_type(path, 0o100000)
# XXX Could also expose Modules/getpath.c:isdir()
def _path_isdir(path):
"""Replacement for os.path.isdir."""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
if not path:
path = _os.getcwd()
return _path_is_mode_type(path, 0o040000)
def _write_atomic(path, data):
"""Best-effort function to write data to a path atomically.
Be prepared to handle a FileExistsError if concurrent writing of the
temporary file is attempted."""
# id() is used to generate a pseudo-random filename.
path_tmp = '{}.{}'.format(path, id(path))
fd = _os.open(path_tmp, _os.O_EXCL | _os.O_CREAT | _os.O_WRONLY, 0o666)
try:
# We first write data to a temporary file, and then use os.replace() to
# perform an atomic rename.
with _io.FileIO(fd, 'wb') as file:
file.write(data)
_os.replace(path_tmp, path)
except OSError:
try:
_os.unlink(path_tmp)
except OSError:
pass
raise
def _wrap(new, old):
"""Simple substitute for functools.wraps."""
for replace in ['__module__', '__name__', '__qualname__', '__doc__']:
if hasattr(old, replace):
setattr(new, replace, getattr(old, replace))
new.__dict__.update(old.__dict__)
_code_type = type(_wrap.__code__)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def new_module(name):
"""Create a new module.
The module is not entered into sys.modules.
"""
2012-05-04 16:46:04 -03:00
return type(_io)(name)
# Module-level locking ########################################################
# A dict mapping module names to weakrefs of _ModuleLock instances
_module_locks = {}
# A dict mapping thread ids to _ModuleLock instances
_blocking_on = {}
class _DeadlockError(RuntimeError):
pass
class _ModuleLock:
"""A recursive lock implementation which is able to detect deadlocks
(e.g. thread 1 trying to take locks A then B, and thread 2 trying to
take locks B then A).
"""
def __init__(self, name):
self.lock = _thread.allocate_lock()
self.wakeup = _thread.allocate_lock()
self.name = name
self.owner = None
self.count = 0
self.waiters = 0
def has_deadlock(self):
# Deadlock avoidance for concurrent circular imports.
me = _thread.get_ident()
tid = self.owner
while True:
lock = _blocking_on.get(tid)
if lock is None:
return False
tid = lock.owner
if tid == me:
return True
def acquire(self):
"""
Acquire the module lock. If a potential deadlock is detected,
a _DeadlockError is raised.
Otherwise, the lock is always acquired and True is returned.
"""
tid = _thread.get_ident()
_blocking_on[tid] = self
try:
while True:
with self.lock:
if self.count == 0 or self.owner == tid:
self.owner = tid
self.count += 1
return True
if self.has_deadlock():
raise _DeadlockError("deadlock detected by %r" % self)
if self.wakeup.acquire(False):
self.waiters += 1
# Wait for a release() call
self.wakeup.acquire()
self.wakeup.release()
finally:
del _blocking_on[tid]
def release(self):
tid = _thread.get_ident()
with self.lock:
if self.owner != tid:
raise RuntimeError("cannot release un-acquired lock")
assert self.count > 0
self.count -= 1
if self.count == 0:
self.owner = None
if self.waiters:
self.waiters -= 1
self.wakeup.release()
def __repr__(self):
return "_ModuleLock(%r) at %d" % (self.name, id(self))
class _DummyModuleLock:
"""A simple _ModuleLock equivalent for Python builds without
multi-threading support."""
def __init__(self, name):
self.name = name
self.count = 0
def acquire(self):
self.count += 1
return True
def release(self):
if self.count == 0:
raise RuntimeError("cannot release un-acquired lock")
self.count -= 1
def __repr__(self):
return "_DummyModuleLock(%r) at %d" % (self.name, id(self))
# The following two functions are for consumption by Python/import.c.
def _get_module_lock(name):
"""Get or create the module lock for a given module name.
Should only be called with the import lock taken."""
lock = None
if name in _module_locks:
lock = _module_locks[name]()
if lock is None:
if _thread is None:
lock = _DummyModuleLock(name)
else:
lock = _ModuleLock(name)
def cb(_):
del _module_locks[name]
_module_locks[name] = _weakref.ref(lock, cb)
return lock
def _lock_unlock_module(name):
"""Release the global import lock, and acquires then release the
module lock for a given module name.
This is used to ensure a module is completely initialized, in the
event it is being imported by another thread.
Should only be called with the import lock taken."""
lock = _get_module_lock(name)
_imp.release_lock()
try:
lock.acquire()
except _DeadlockError:
# Concurrent circular import, we'll accept a partially initialized
# module object.
pass
else:
lock.release()
# Finder/loader utility code ##################################################
_PYCACHE = '__pycache__'
SOURCE_SUFFIXES = ['.py'] # _setup() adds .pyw as needed.
DEBUG_BYTECODE_SUFFIXES = ['.pyc']
OPTIMIZED_BYTECODE_SUFFIXES = ['.pyo']
if __debug__:
BYTECODE_SUFFIXES = DEBUG_BYTECODE_SUFFIXES
else:
BYTECODE_SUFFIXES = OPTIMIZED_BYTECODE_SUFFIXES
def cache_from_source(path, debug_override=None):
"""Given the path to a .py file, return the path to its .pyc/.pyo file.
The .py file does not need to exist; this simply returns the path to the
.pyc/.pyo file calculated as if the .py file were imported. The extension
will be .pyc unless __debug__ is not defined, then it will be .pyo.
If debug_override is not None, then it must be a boolean and is taken as
the value of __debug__ instead.
"""
debug = __debug__ if debug_override is None else debug_override
if debug:
suffixes = DEBUG_BYTECODE_SUFFIXES
else:
suffixes = OPTIMIZED_BYTECODE_SUFFIXES
head, tail = _path_split(path)
base_filename, sep, _ = tail.partition('.')
filename = ''.join([base_filename, sep, _TAG, suffixes[0]])
return _path_join(head, _PYCACHE, filename)
def _verbose_message(message, *args):
"""Print the message to stderr if -v/PYTHONVERBOSE is turned on."""
if sys.flags.verbose:
2012-04-15 16:21:32 -03:00
if not message.startswith(('#', 'import ')):
message = '# ' + message
print(message.format(*args), file=sys.stderr)
def set_package(fxn):
"""Set __package__ on the returned module."""
def set_package_wrapper(*args, **kwargs):
module = fxn(*args, **kwargs)
if not hasattr(module, '__package__') or module.__package__ is None:
module.__package__ = module.__name__
if not hasattr(module, '__path__'):
module.__package__ = module.__package__.rpartition('.')[0]
return module
_wrap(set_package_wrapper, fxn)
return set_package_wrapper
def set_loader(fxn):
"""Set __loader__ on the returned module."""
def set_loader_wrapper(self, *args, **kwargs):
module = fxn(self, *args, **kwargs)
if not hasattr(module, '__loader__'):
module.__loader__ = self
return module
_wrap(set_loader_wrapper, fxn)
return set_loader_wrapper
def module_for_loader(fxn):
"""Decorator to handle selecting the proper module for loaders.
The decorated function is passed the module to use instead of the module
name. The module passed in to the function is either from sys.modules if
it already exists or is a new module. If the module is new, then __name__
is set the first argument to the method, __loader__ is set to self, and
__package__ is set accordingly (if self.is_package() is defined) will be set
before it is passed to the decorated function (if self.is_package() does
not work for the module it will be set post-load).
If an exception is raised and the decorator created the module it is
subsequently removed from sys.modules.
The decorator assumes that the decorated function takes the module name as
the second argument.
"""
def module_for_loader_wrapper(self, fullname, *args, **kwargs):
module = sys.modules.get(fullname)
is_reload = module is not None
if not is_reload:
# This must be done before open() is called as the 'io' module
# implicitly imports 'locale' and would otherwise trigger an
# infinite loop.
module = new_module(fullname)
sys.modules[fullname] = module
module.__loader__ = self
try:
is_package = self.is_package(fullname)
except (ImportError, AttributeError):
pass
else:
if is_package:
module.__package__ = fullname
else:
module.__package__ = fullname.rpartition('.')[0]
try:
module.__initializing__ = True
# If __package__ was not set above, __import__() will do it later.
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return fxn(self, module, *args, **kwargs)
except:
if not is_reload:
del sys.modules[fullname]
raise
finally:
module.__initializing__ = False
_wrap(module_for_loader_wrapper, fxn)
return module_for_loader_wrapper
def _check_name(method):
"""Decorator to verify that the module being requested matches the one the
loader can handle.
The first argument (self) must define _name which the second argument is
compared against. If the comparison fails then ImportError is raised.
"""
def _check_name_wrapper(self, name=None, *args, **kwargs):
if name is None:
name = self.name
elif self.name != name:
raise ImportError("loader cannot handle %s" % name, name=name)
return method(self, name, *args, **kwargs)
_wrap(_check_name_wrapper, method)
return _check_name_wrapper
def _requires_builtin(fxn):
"""Decorator to verify the named module is built-in."""
def _requires_builtin_wrapper(self, fullname):
if fullname not in sys.builtin_module_names:
raise ImportError("{0} is not a built-in module".format(fullname),
name=fullname)
return fxn(self, fullname)
_wrap(_requires_builtin_wrapper, fxn)
return _requires_builtin_wrapper
def _requires_frozen(fxn):
"""Decorator to verify the named module is frozen."""
def _requires_frozen_wrapper(self, fullname):
if not _imp.is_frozen(fullname):
raise ImportError("{0} is not a frozen module".format(fullname),
name=fullname)
return fxn(self, fullname)
_wrap(_requires_frozen_wrapper, fxn)
return _requires_frozen_wrapper
# Loaders #####################################################################
class BuiltinImporter:
"""Meta path import for built-in modules.
All methods are either class or static methods to avoid the need to
instantiate the class.
"""
@classmethod
def module_repr(cls, module):
return "<module '{}' (built-in)>".format(module.__name__)
@classmethod
def find_module(cls, fullname, path=None):
"""Find the built-in module.
If 'path' is ever specified then the search is considered a failure.
"""
if path is not None:
return None
return cls if _imp.is_builtin(fullname) else None
@classmethod
@set_package
@set_loader
@_requires_builtin
def load_module(cls, fullname):
"""Load a built-in module."""
is_reload = fullname in sys.modules
try:
return _imp.init_builtin(fullname)
except:
if not is_reload and fullname in sys.modules:
del sys.modules[fullname]
raise
@classmethod
@_requires_builtin
def get_code(cls, fullname):
"""Return None as built-in modules do not have code objects."""
return None
@classmethod
@_requires_builtin
def get_source(cls, fullname):
"""Return None as built-in modules do not have source code."""
return None
@classmethod
@_requires_builtin
def is_package(cls, fullname):
"""Return None as built-in modules are never packages."""
return False
class FrozenImporter:
"""Meta path import for frozen modules.
All methods are either class or static methods to avoid the need to
instantiate the class.
"""
@classmethod
def module_repr(cls, m):
return "<module '{}' (frozen)>".format(m.__name__)
@classmethod
def find_module(cls, fullname, path=None):
"""Find a frozen module."""
return cls if _imp.is_frozen(fullname) else None
@classmethod
@set_package
@set_loader
@_requires_frozen
def load_module(cls, fullname):
"""Load a frozen module."""
is_reload = fullname in sys.modules
try:
m = _imp.init_frozen(fullname)
# Let our own module_repr() method produce a suitable repr.
del m.__file__
return m
except:
if not is_reload and fullname in sys.modules:
del sys.modules[fullname]
raise
@classmethod
@_requires_frozen
def get_code(cls, fullname):
"""Return the code object for the frozen module."""
return _imp.get_frozen_object(fullname)
@classmethod
@_requires_frozen
def get_source(cls, fullname):
"""Return None as frozen modules do not have source code."""
return None
@classmethod
@_requires_frozen
def is_package(cls, fullname):
"""Return if the frozen module is a package."""
return _imp.is_frozen_package(fullname)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
class _LoaderBasics:
"""Base class of common code needed by both SourceLoader and
SourcelessFileLoader."""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def is_package(self, fullname):
"""Concrete implementation of InspectLoader.is_package by checking if
the path returned by get_filename has a filename of '__init__.py'."""
filename = _path_split(self.get_filename(fullname))[1]
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return filename.rsplit('.', 1)[0] == '__init__'
def _bytes_from_bytecode(self, fullname, data, bytecode_path, source_stats):
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""Return the marshalled bytes from bytecode, verifying the magic
number, timestamp and source size along the way.
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
If source_stats is None then skip the timestamp check.
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""
magic = data[:4]
raw_timestamp = data[4:8]
raw_size = data[8:12]
if magic != _MAGIC_NUMBER:
msg = 'bad magic number in {!r}: {!r}'.format(fullname, magic)
raise ImportError(msg, name=fullname, path=bytecode_path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
elif len(raw_timestamp) != 4:
message = 'bad timestamp in {}'.format(fullname)
_verbose_message(message)
raise EOFError(message)
elif len(raw_size) != 4:
message = 'bad size in {}'.format(fullname)
_verbose_message(message)
raise EOFError(message)
if source_stats is not None:
try:
source_mtime = int(source_stats['mtime'])
except KeyError:
pass
else:
if _r_long(raw_timestamp) != source_mtime:
message = 'bytecode is stale for {}'.format(fullname)
_verbose_message(message)
raise ImportError(message, name=fullname,
path=bytecode_path)
try:
source_size = source_stats['size'] & 0xFFFFFFFF
except KeyError:
pass
else:
if _r_long(raw_size) != source_size:
2012-02-25 01:48:17 -04:00
raise ImportError(
"bytecode is stale for {}".format(fullname),
name=fullname, path=bytecode_path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
# Can't return the code object as errors from marshal loading need to
# propagate even when source is available.
return data[12:]
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
@module_for_loader
def _load_module(self, module, *, sourceless=False):
"""Helper for load_module able to handle either source or sourceless
loading."""
name = module.__name__
code_object = self.get_code(name)
module.__file__ = self.get_filename(name)
if not sourceless:
module.__cached__ = cache_from_source(module.__file__)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
else:
module.__cached__ = module.__file__
module.__package__ = name
if self.is_package(name):
module.__path__ = [_path_split(module.__file__)[0]]
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
else:
module.__package__ = module.__package__.rpartition('.')[0]
module.__loader__ = self
exec(code_object, module.__dict__)
return module
class SourceLoader(_LoaderBasics):
def path_mtime(self, path):
"""Optional method that returns the modification time (an int) for the
specified path, where path is a str.
"""
raise NotImplementedError
def path_stats(self, path):
"""Optional method returning a metadata dict for the specified path
to by the path (str).
Possible keys:
- 'mtime' (mandatory) is the numeric timestamp of last source
code modification;
- 'size' (optional) is the size in bytes of the source code.
Implementing this method allows the loader to read bytecode files.
"""
return {'mtime': self.path_mtime(path)}
def set_data(self, path, data):
"""Optional method which writes data (bytes) to a file path (a str).
Implementing this method allows for the writing of bytecode files.
"""
raise NotImplementedError
def get_source(self, fullname):
"""Concrete implementation of InspectLoader.get_source."""
import tokenize
path = self.get_filename(fullname)
try:
source_bytes = self.get_data(path)
except IOError:
raise ImportError("source not available through get_data()",
name=fullname)
encoding = tokenize.detect_encoding(_io.BytesIO(source_bytes).readline)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 19:32:41 -03:00
newline_decoder = _io.IncrementalNewlineDecoder(None, True)
return newline_decoder.decode(source_bytes.decode(encoding[0]))
def get_code(self, fullname):
"""Concrete implementation of InspectLoader.get_code.
Reading of bytecode requires path_stats to be implemented. To write
bytecode, set_data must also be implemented.
"""
source_path = self.get_filename(fullname)
bytecode_path = cache_from_source(source_path)
source_mtime = None
if bytecode_path is not None:
try:
st = self.path_stats(source_path)
except NotImplementedError:
pass
else:
source_mtime = int(st['mtime'])
try:
data = self.get_data(bytecode_path)
except IOError:
pass
else:
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
try:
bytes_data = self._bytes_from_bytecode(fullname, data,
bytecode_path,
st)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
except (ImportError, EOFError):
pass
else:
_verbose_message('{} matches {}', bytecode_path,
source_path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
found = marshal.loads(bytes_data)
if isinstance(found, _code_type):
_imp._fix_co_filename(found, source_path)
_verbose_message('code object from {}',
bytecode_path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return found
else:
msg = "Non-code object in {}"
raise ImportError(msg.format(bytecode_path),
name=fullname, path=bytecode_path)
source_bytes = self.get_data(source_path)
code_object = compile(source_bytes, source_path, 'exec',
dont_inherit=True)
_verbose_message('code object from {}', source_path)
if (not sys.dont_write_bytecode and bytecode_path is not None and
2012-02-25 01:48:17 -04:00
source_mtime is not None):
data = bytearray(_MAGIC_NUMBER)
data.extend(_w_long(source_mtime))
data.extend(_w_long(len(source_bytes)))
data.extend(marshal.dumps(code_object))
try:
self.set_data(bytecode_path, data)
_verbose_message('wrote {!r}', bytecode_path)
except NotImplementedError:
pass
return code_object
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def load_module(self, fullname):
"""Concrete implementation of Loader.load_module.
Requires ExecutionLoader.get_filename and ResourceLoader.get_data to be
implemented to load source code. Use of bytecode is dictated by whether
get_code uses/writes bytecode.
"""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return self._load_module(fullname)
class FileLoader:
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""Base file loader class which implements the loader protocol methods that
require file system usage."""
def __init__(self, fullname, path):
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""Cache the module name and the path to the file found by the
finder."""
self.name = fullname
self.path = path
@_check_name
def load_module(self, fullname):
"""Load a module from a file."""
# Issue #14857: Avoid the zero-argument form so the implementation
# of that form can be updated without breaking the frozen module
return super(FileLoader, self).load_module(fullname)
@_check_name
def get_filename(self, fullname):
"""Return the path to the source file as found by the finder."""
return self.path
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def get_data(self, path):
"""Return the data from path as raw bytes."""
2010-09-03 16:55:26 -03:00
with _io.FileIO(path, 'r') as file:
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return file.read()
class SourceFileLoader(FileLoader, SourceLoader):
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""Concrete implementation of SourceLoader using the file system."""
def path_stats(self, path):
"""Return the metadat for the path."""
st = _os.stat(path)
return {'mtime': st.st_mtime, 'size': st.st_size}
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def set_data(self, path, data):
"""Write bytes data to a file."""
parent, filename = _path_split(path)
path_parts = []
# Figure out what directories are missing.
while parent and not _path_isdir(parent):
parent, part = _path_split(parent)
path_parts.append(part)
# Create needed directories.
for part in reversed(path_parts):
parent = _path_join(parent, part)
try:
_os.mkdir(parent)
except FileExistsError:
# Probably another Python process already created the dir.
continue
except PermissionError:
# If can't get proper access, then just forget about writing
# the data.
return
try:
_write_atomic(path, data)
_verbose_message('created {!r}', path)
except (PermissionError, FileExistsError):
# Don't worry if you can't write bytecode or someone is writing
# it at the same time.
pass
class SourcelessFileLoader(FileLoader, _LoaderBasics):
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""Loader which handles sourceless file imports."""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def load_module(self, fullname):
return self._load_module(fullname, sourceless=True)
def get_code(self, fullname):
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
path = self.get_filename(fullname)
data = self.get_data(path)
bytes_data = self._bytes_from_bytecode(fullname, data, path, None)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
found = marshal.loads(bytes_data)
if isinstance(found, _code_type):
_verbose_message('code object from {!r}', path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
return found
else:
raise ImportError("Non-code object in {}".format(path),
name=fullname, path=path)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def get_source(self, fullname):
"""Return None as there is no source code."""
return None
class ExtensionFileLoader:
"""Loader for extension modules.
The constructor is designed to work with FileFinder.
"""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def __init__(self, name, path):
self.name = name
self.path = path
@_check_name
@set_package
@set_loader
def load_module(self, fullname):
"""Load an extension module."""
is_reload = fullname in sys.modules
try:
module = _imp.load_dynamic(fullname, self.path)
_verbose_message('extension module loaded from {!r}', self.path)
return module
except:
if not is_reload and fullname in sys.modules:
del sys.modules[fullname]
raise
def is_package(self, fullname):
"""Return False as an extension module can never be a package."""
return False
def get_code(self, fullname):
"""Return None as an extension module cannot create a code object."""
return None
def get_source(self, fullname):
"""Return None as extension modules have no source code."""
return None
class _NamespacePath:
"""Represents a namespace package's path. It uses the module name
to find its parent module, and from there it looks up the parent's
__path__. When this changes, the module's own path is recomputed,
using path_finder. For top-leve modules, the parent module's path
is sys.path."""
def __init__(self, name, path, path_finder):
self._name = name
self._path = path
self._last_parent_path = tuple(self._get_parent_path())
self._path_finder = path_finder
def _find_parent_path_names(self):
"""Returns a tuple of (parent-module-name, parent-path-attr-name)"""
parent, dot, me = self._name.rpartition('.')
if dot == '':
# This is a top-level module. sys.path contains the parent path.
return 'sys', 'path'
# Not a top-level module. parent-module.__path__ contains the
# parent path.
return parent, '__path__'
def _get_parent_path(self):
parent_module_name, path_attr_name = self._find_parent_path_names()
return getattr(sys.modules[parent_module_name], path_attr_name)
def _recalculate(self):
# If the parent's path has changed, recalculate _path
parent_path = tuple(self._get_parent_path()) # Make a copy
if parent_path != self._last_parent_path:
loader, new_path = self._path_finder(self._name, parent_path)
# Note that no changes are made if a loader is returned, but we
# do remember the new parent path
if loader is None:
self._path = new_path
self._last_parent_path = parent_path # Save the copy
return self._path
def __iter__(self):
return iter(self._recalculate())
def __len__(self):
return len(self._recalculate())
def __repr__(self):
return "_NamespacePath({0!r})".format(self._path)
def __contains__(self, item):
return item in self._recalculate()
def append(self, item):
self._path.append(item)
class NamespaceLoader:
def __init__(self, name, path, path_finder):
self._path = _NamespacePath(name, path, path_finder)
@classmethod
def module_repr(cls, module):
return "<module '{}' (namespace)>".format(module.__name__)
@module_for_loader
def load_module(self, module):
"""Load a namespace module."""
_verbose_message('namespace module loaded with path {!r}', self._path)
module.__path__ = self._path
return module
# Finders #####################################################################
class PathFinder:
"""Meta path finder for sys.(path|path_hooks|path_importer_cache)."""
@classmethod
def _path_hooks(cls, path):
"""Search sequence of hooks for a finder for 'path'.
If 'hooks' is false then use sys.path_hooks.
"""
if not sys.path_hooks:
_warnings.warn('sys.path_hooks is empty', ImportWarning)
for hook in sys.path_hooks:
try:
return hook(path)
except ImportError:
continue
else:
return None
@classmethod
def _path_importer_cache(cls, path):
"""Get the finder for the path from sys.path_importer_cache.
If the path is not in the cache, find the appropriate finder and cache
it. If no finder is available, store None.
"""
if path == '':
path = '.'
try:
finder = sys.path_importer_cache[path]
except KeyError:
finder = cls._path_hooks(path)
sys.path_importer_cache[path] = finder
return finder
@classmethod
def _get_loader(cls, fullname, path):
"""Find the loader or namespace_path for this module/package name."""
# If this ends up being a namespace package, namespace_path is
# the list of paths that will become its __path__
namespace_path = []
for entry in path:
finder = cls._path_importer_cache(entry)
if finder is not None:
if hasattr(finder, 'find_loader'):
loader, portions = finder.find_loader(fullname)
else:
loader = finder.find_module(fullname)
portions = []
if loader is not None:
# We found a loader: return it immediately.
return (loader, namespace_path)
# This is possibly part of a namespace package.
# Remember these path entries (if any) for when we
# create a namespace package, and continue iterating
# on path.
namespace_path.extend(portions)
else:
return (None, namespace_path)
@classmethod
def find_module(cls, fullname, path=None):
"""Find the module on sys.path or 'path' based on sys.path_hooks and
sys.path_importer_cache."""
if path is None:
path = sys.path
loader, namespace_path = cls._get_loader(fullname, path)
if loader is not None:
return loader
else:
if namespace_path:
# We found at least one namespace path. Return a
# loader which can create the namespace package.
return NamespaceLoader(fullname, namespace_path, cls._get_loader)
else:
return None
class FileFinder:
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
"""File-based finder.
Interactions with the file system are cached for performance, being
refreshed when the directory the finder is handling has been modified.
"""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def __init__(self, path, *details):
"""Initialize with the path to search on and a variable number of
3-tuples containing the loader, file suffixes the loader recognizes,
and a boolean of whether the loader handles packages."""
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
packages = []
modules = []
for loader, suffixes, supports_packages in details:
modules.extend((suffix, loader) for suffix in suffixes)
if supports_packages:
packages.extend((suffix, loader) for suffix in suffixes)
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
self.packages = packages
self.modules = modules
# Base (directory) path
self.path = path or '.'
self._path_mtime = -1
self._path_cache = set()
self._relaxed_path_cache = set()
def invalidate_caches(self):
"""Invalidate the directory mtime."""
self._path_mtime = -1
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
def find_module(self, fullname):
"""Try to find a loader for the specified module."""
# Call find_loader(). If it returns a string (indicating this
# is a namespace package portion), generate a warning and
# return None.
loader, portions = self.find_loader(fullname)
assert len(portions) in [0, 1]
if loader is None and len(portions):
msg = "Not importing directory {}: missing __init__"
_warnings.warn(msg.format(portions[0]), ImportWarning)
return loader
def find_loader(self, fullname):
"""Try to find a loader for the specified module, or the namespace
package portions. Returns (loader, list-of-portions)."""
tail_module = fullname.rpartition('.')[2]
try:
mtime = _os.stat(self.path).st_mtime
except OSError:
mtime = -1
if mtime != self._path_mtime:
self._fill_cache()
self._path_mtime = mtime
# tail_module keeps the original casing, for __file__ and friends
if _relax_case():
cache = self._relaxed_path_cache
cache_module = tail_module.lower()
else:
cache = self._path_cache
cache_module = tail_module
2012-04-14 19:37:07 -03:00
# Check if the module is the name of a directory (and thus a package).
if cache_module in cache:
base_path = _path_join(self.path, tail_module)
if _path_isdir(base_path):
for suffix, loader in self.packages:
init_filename = '__init__' + suffix
full_path = _path_join(base_path, init_filename)
if _path_isfile(full_path):
return (loader(fullname, full_path), [base_path])
else:
# A namespace package, return the path
return (None, [base_path])
2012-04-14 19:37:07 -03:00
# Check for a file w/ a proper suffix exists.
Make importlib.abc.SourceLoader the primary mechanism for importlib. This required moving the class from importlib/abc.py into importlib/_bootstrap.py and jiggering some code to work better with the class. This included changing how the file finder worked to better meet import semantics. This also led to fixing importlib to handle the empty string from sys.path as import currently does (and making me wish we didn't support that instead just required people to insert '.' instead to represent cwd). It also required making the new set_data abstractmethod create any needed subdirectories implicitly thanks to __pycache__ (it was either this or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir method or have set_data with no data arg mean to create a directory). Lastly, as an optimization the file loaders cache the file path where the finder found something to use for loading (this is thanks to having a sourceless loader separate from the source loader to simplify the code and cut out stat calls). Unfortunately test_runpy assumed a loader would always work for a module, even if you changed from underneath it what it was expected to work with. By simply dropping the previous loader in test_runpy so the proper loader can be returned by the finder fixed the failure. At this point importlib deviates from import on two points: 1. The exception raised when trying to import a file is different (import does an explicit file check to print a special message, importlib just says the path cannot be imported as if it was just some module name). 2. the co_filename on a code object is not being set to where bytecode was actually loaded from instead of where the marshalled code object originally came from (a solution for this has already been agreed upon on python-dev but has not been implemented yet; issue8611).
2010-07-03 18:48:25 -03:00
for suffix, loader in self.modules:
if cache_module + suffix in cache:
full_path = _path_join(self.path, tail_module + suffix)
if _path_isfile(full_path):
return (loader(fullname, full_path), [])
return (None, [])
def _fill_cache(self):
"""Fill the cache of potential modules and packages for this directory."""
path = self.path
contents = _os.listdir(path)
# We store two cached versions, to handle runtime changes of the
# PYTHONCASEOK environment variable.
if not sys.platform.startswith('win'):
self._path_cache = set(contents)
else:
# Windows users can import modules with case-insensitive file
# suffixes (for legacy reasons). Make the suffix lowercase here
# so it's done once instead of for every import. This is safe as
# the specified suffixes to check against are always specified in a
# case-sensitive manner.
lower_suffix_contents = set()
for item in contents:
name, dot, suffix = item.partition('.')
if dot:
new_name = '{}.{}'.format(name, suffix.lower())
else:
new_name = name
lower_suffix_contents.add(new_name)
self._path_cache = lower_suffix_contents
if sys.platform.startswith(_CASE_INSENSITIVE_PLATFORMS):
self._relaxed_path_cache = set(fn.lower() for fn in contents)
@classmethod
def path_hook(cls, *loader_details):
"""A class method which returns a closure to use on sys.path_hook
which will return an instance using the specified loaders and the path
called on the closure.
If the path called on the closure is not a directory, ImportError is
raised.
"""
def path_hook_for_FileFinder(path):
"""Path hook for importlib.machinery.FileFinder."""
if not _path_isdir(path):
raise ImportError("only directories are supported", path=path)
return cls(path, *loader_details)
return path_hook_for_FileFinder
# Import itself ###############################################################
class _ImportLockContext:
"""Context manager for the import lock."""
def __enter__(self):
"""Acquire the import lock."""
_imp.acquire_lock()
def __exit__(self, exc_type, exc_value, exc_traceback):
"""Release the import lock regardless of any raised exceptions."""
_imp.release_lock()
def _resolve_name(name, package, level):
"""Resolve a relative module name to an absolute one."""
2012-02-25 01:48:17 -04:00
bits = package.rsplit('.', level - 1)
2012-02-24 12:20:54 -04:00
if len(bits) < level:
raise ValueError('attempted relative import beyond top-level package')
base = bits[0]
return '{0}.{1}'.format(base, name) if name else base
def _find_module(name, path):
"""Find a module's loader."""
if not sys.meta_path:
_warnings.warn('sys.meta_path is empty', ImportWarning)
for finder in sys.meta_path:
with _ImportLockContext():
loader = finder.find_module(name, path)
if loader is not None:
# The parent import may have already imported this module.
if name not in sys.modules:
return loader
else:
return sys.modules[name].__loader__
else:
return None
def _sanity_check(name, package, level):
"""Verify arguments are "sane"."""
if not isinstance(name, str):
raise TypeError("module name must be str, not {}".format(type(name)))
if level < 0:
raise ValueError('level must be >= 0')
if package:
if not isinstance(package, str):
raise TypeError("__package__ not set to a string")
elif package not in sys.modules:
msg = ("Parent module {0!r} not loaded, cannot perform relative "
"import")
raise SystemError(msg.format(package))
if not name and level == 0:
raise ValueError("Empty module name")
_ERR_MSG = 'No module named {!r}'
def _find_and_load_unlocked(name, import_):
path = None
parent = name.rpartition('.')[0]
if parent:
if parent not in sys.modules:
import_(parent)
# Crazy side-effects!
if name in sys.modules:
return sys.modules[name]
# Backwards-compatibility; be nicer to skip the dict lookup.
parent_module = sys.modules[parent]
try:
path = parent_module.__path__
except AttributeError:
msg = (_ERR_MSG + '; {} is not a package').format(name, parent)
raise ImportError(msg, name=name)
loader = _find_module(name, path)
if loader is None:
raise ImportError(_ERR_MSG.format(name), name=name)
elif name not in sys.modules:
# The parent import may have already imported this module.
loader.load_module(name)
_verbose_message('import {!r} # {!r}', name, loader)
# Backwards-compatibility; be nicer to skip the dict lookup.
module = sys.modules[name]
if parent:
# Set the module as an attribute on its parent.
parent_module = sys.modules[parent]
setattr(parent_module, name.rpartition('.')[2], module)
# Set __package__ if the loader did not.
if not hasattr(module, '__package__') or module.__package__ is None:
try:
module.__package__ = module.__name__
if not hasattr(module, '__path__'):
module.__package__ = module.__package__.rpartition('.')[0]
except AttributeError:
pass
# Set loader if need be.
if not hasattr(module, '__loader__'):
try:
module.__loader__ = loader
except AttributeError:
pass
return module
def _find_and_load(name, import_):
"""Find and load the module, and release the import lock."""
try:
lock = _get_module_lock(name)
finally:
_imp.release_lock()
lock.acquire()
try:
return _find_and_load_unlocked(name, import_)
finally:
lock.release()
def _gcd_import(name, package=None, level=0):
"""Import and return the module based on its name, the package the call is
being made from, and the level adjustment.
This function represents the greatest common denominator of functionality
between import_module and __import__. This includes setting __package__ if
the loader did not.
"""
_sanity_check(name, package, level)
if level > 0:
name = _resolve_name(name, package, level)
_imp.acquire_lock()
if name not in sys.modules:
return _find_and_load(name, _gcd_import)
module = sys.modules[name]
if module is None:
_imp.release_lock()
message = ("import of {} halted; "
"None in sys.modules".format(name))
raise ImportError(message, name=name)
_lock_unlock_module(name)
return module
def _handle_fromlist(module, fromlist, import_):
"""Figure out what __import__ should return.
The import_ parameter is a callable which takes the name of module to
import. It is required to decouple the function from assuming importlib's
import implementation is desired.
"""
# The hell that is fromlist ...
# If a package was imported, try to import stuff from fromlist.
if hasattr(module, '__path__'):
if '*' in fromlist and hasattr(module, '__all__'):
fromlist = list(fromlist)
fromlist.remove('*')
fromlist.extend(module.__all__)
2012-02-25 01:48:17 -04:00
for x in (y for y in fromlist if not hasattr(module, y)):
try:
import_('{0}.{1}'.format(module.__name__, x))
except ImportError:
pass
return module
def _calc___package__(globals):
"""Calculate what __package__ should be.
__package__ is not guaranteed to be defined or could be set to None
to represent that its proper value is unknown.
"""
package = globals.get('__package__')
if package is None:
package = globals['__name__']
if '__path__' not in globals:
package = package.rpartition('.')[0]
return package
def __import__(name, globals={}, locals={}, fromlist=[], level=0):
"""Import a module.
The 'globals' argument is used to infer where the import is occuring from
to handle relative imports. The 'locals' argument is ignored. The
'fromlist' argument specifies what should exist as attributes on the module
being imported (e.g. ``from module import <fromlist>``). The 'level'
argument represents the package location to import from in a relative
import (e.g. ``from ..pkg import mod`` would have a 'level' of 2).
"""
if level == 0:
module = _gcd_import(name)
else:
package = _calc___package__(globals)
module = _gcd_import(name, package, level)
if not fromlist:
# Return up to the first dot in 'name'. This is complicated by the fact
# that 'name' may be relative.
if level == 0:
return _gcd_import(name.partition('.')[0])
elif not name:
return module
else:
cut_off = len(name) - len(name.partition('.')[0])
return sys.modules[module.__name__[:len(module.__name__)-cut_off]]
else:
return _handle_fromlist(module, fromlist, _gcd_import)
_MAGIC_NUMBER = None # Set in _setup()
_TAG = None # Set in _setup()
def _setup(sys_module, _imp_module):
"""Setup importlib by importing needed built-in modules and injecting them
into the global namespace.
As sys is needed for sys.modules access and _imp is needed to load built-in
modules, those two modules must be explicitly passed in.
"""
global _imp, sys
_imp = _imp_module
sys = sys_module
for module in (_imp, sys):
if not hasattr(module, '__loader__'):
module.__loader__ = BuiltinImporter
self_module = sys.modules[__name__]
for builtin_name in ('_io', '_warnings', 'builtins', 'marshal'):
if builtin_name not in sys.modules:
builtin_module = BuiltinImporter.load_module(builtin_name)
else:
builtin_module = sys.modules[builtin_name]
setattr(self_module, builtin_name, builtin_module)
os_details = ('posix', ['/']), ('nt', ['\\', '/']), ('os2', ['\\', '/'])
for builtin_os, path_separators in os_details:
# Assumption made in _path_join()
assert all(len(sep) == 1 for sep in path_separators)
path_sep = path_separators[0]
if builtin_os in sys.modules:
os_module = sys.modules[builtin_os]
break
else:
try:
os_module = BuiltinImporter.load_module(builtin_os)
# TODO: rip out os2 code after 3.3 is released as per PEP 11
if builtin_os == 'os2' and 'EMX GCC' in sys.version:
path_sep = path_separators[1]
break
except ImportError:
continue
else:
raise ImportError('importlib requires posix or nt')
try:
thread_module = BuiltinImporter.load_module('_thread')
except ImportError:
# Python was built without threads
thread_module = None
weakref_module = BuiltinImporter.load_module('_weakref')
setattr(self_module, '_os', os_module)
setattr(self_module, '_thread', thread_module)
setattr(self_module, '_weakref', weakref_module)
setattr(self_module, 'path_sep', path_sep)
setattr(self_module, 'path_separators', set(path_separators))
# Constants
2012-02-20 08:52:47 -04:00
setattr(self_module, '_relax_case', _make_relax_case())
setattr(self_module, '_MAGIC_NUMBER', _imp_module.get_magic())
setattr(self_module, '_TAG', _imp.get_tag())
if builtin_os == 'nt':
SOURCE_SUFFIXES.append('.pyw')
def _install(sys_module, _imp_module):
2012-05-04 17:47:54 -03:00
"""Install importlib as the implementation of import."""
_setup(sys_module, _imp_module)
extensions = ExtensionFileLoader, _imp_module.extension_suffixes(), False
source = SourceFileLoader, SOURCE_SUFFIXES, True
bytecode = SourcelessFileLoader, BYTECODE_SUFFIXES, True
supported_loaders = [extensions, source, bytecode]
sys.path_hooks.extend([FileFinder.path_hook(*supported_loaders)])
sys.meta_path.extend([BuiltinImporter, FrozenImporter, PathFinder])