_bootstrap.py

"""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'


def _make_relax_case():
    if sys.platform.startswith(_CASE_INSENSITIVE_PLATFORMS):
        def _relax_case():
            """True if filenames must be checked case-insensitively."""
            return b'PYTHONCASEOK' in _os.environ
    else:
        def _relax_case():
            """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."""
    if not path:
        path = _os.getcwd()
    return _path_is_mode_type(path, 0o040000)


def _write_atomic(path, data, mode=0o666):
    """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, mode & 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.update_wrapper."""
    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__)


def new_module(name):
    """Create a new module.

    The module is not entered into sys.modules.

    """
    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
    try:
        lock = _module_locks[name]()
    except KeyError:
        pass
    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()

# Frame stripping magic ###############################################

def _call_with_frames_removed(f, *args, **kwds):
    """remove_importlib_frames in import.c will always remove sequences
    of importlib frames that end with a call to this function

    Use it instead of a normal call in places where including the importlib
    frames introduces unwanted noise into the traceback (e.g. when executing
    module code)
    """
    return f(*args, **kwds)


# Finder/loader utility code ###############################################

"""Magic word to reject .pyc files generated by other Python versions.
It should change for each incompatible change to the bytecode.

The value of CR and LF is incorporated so if you ever read or write
a .pyc file in text mode the magic number will be wrong; also, the
Apple MPW compiler swaps their values, botching string constants.

The magic numbers must be spaced apart at least 2 values, as the
-U interpeter flag will cause MAGIC+1 being used. They have been
odd numbers for some time now.

There were a variety of old schemes for setting the magic number.
The current working scheme is to increment the previous value by
10.

Starting with the adoption of PEP 3147 in Python 3.2, every bump in magic
number also includes a new "magic tag", i.e. a human readable string used
to represent the magic number in __pycache__ directories.  When you change
the magic number, you must also set a new unique magic tag.  Generally this
can be named after the Python major version of the magic number bump, but
it can really be anything, as long as it's different than anything else
that's come before.  The tags are included in the following table, starting
with Python 3.2a0.

Known values:
 Python 1.5:   20121
 Python 1.5.1: 20121
    Python 1.5.2: 20121
    Python 1.6:   50428
    Python 2.0:   50823
    Python 2.0.1: 50823
    Python 2.1:   60202
    Python 2.1.1: 60202
    Python 2.1.2: 60202
    Python 2.2:   60717
    Python 2.3a0: 62011
    Python 2.3a0: 62021
    Python 2.3a0: 62011 (!)
    Python 2.4a0: 62041
    Python 2.4a3: 62051
    Python 2.4b1: 62061
    Python 2.5a0: 62071
    Python 2.5a0: 62081 (ast-branch)
    Python 2.5a0: 62091 (with)
    Python 2.5a0: 62092 (changed WITH_CLEANUP opcode)
    Python 2.5b3: 62101 (fix wrong code: for x, in ...)
    Python 2.5b3: 62111 (fix wrong code: x += yield)
    Python 2.5c1: 62121 (fix wrong lnotab with for loops and
                         storing constants that should have been removed)
    Python 2.5c2: 62131 (fix wrong code: for x, in ... in listcomp/genexp)
    Python 2.6a0: 62151 (peephole optimizations and STORE_MAP opcode)
    Python 2.6a1: 62161 (WITH_CLEANUP optimization)
    Python 3000:   3000
                   3010 (removed UNARY_CONVERT)
                   3020 (added BUILD_SET)
                   3030 (added keyword-only parameters)
                   3040 (added signature annotations)
                   3050 (print becomes a function)
                   3060 (PEP 3115 metaclass syntax)
                   3061 (string literals become unicode)
                   3071 (PEP 3109 raise changes)
                   3081 (PEP 3137 make __file__ and __name__ unicode)
                   3091 (kill str8 interning)
                   3101 (merge from 2.6a0, see 62151)
                   3103 (__file__ points to source file)
    Python 3.0a4: 3111 (WITH_CLEANUP optimization).
    Python 3.0a5: 3131 (lexical exception stacking, including POP_EXCEPT)
    Python 3.1a0: 3141 (optimize list, set and dict comprehensions:
            change LIST_APPEND and SET_ADD, add MAP_ADD)
    Python 3.1a0: 3151 (optimize conditional branches:
            introduce POP_JUMP_IF_FALSE and POP_JUMP_IF_TRUE)
    Python 3.2a0: 3160 (add SETUP_WITH)
                  tag: cpython-32
    Python 3.2a1: 3170 (add DUP_TOP_TWO, remove DUP_TOPX and ROT_FOUR)
                  tag: cpython-32
    Python 3.2a2  3180 (add DELETE_DEREF)
    Python 3.3a0  3190 __class__ super closure changed
    Python 3.3a0  3200 (__qualname__ added)
                     3210 (added size modulo 2**32 to the pyc header)
    Python 3.3a1  3220 (changed PEP 380 implementation)
    Python 3.3a4  3230 (revert changes to implicit __class__ closure)

MAGIC must change whenever the bytecode emitted by the compiler may no
longer be understood by older implementations of the eval loop (usually
due to the addition of new opcodes).

"""
_RAW_MAGIC_NUMBER = 3230 | ord('\r') << 16 | ord('\n') << 24
_MAGIC_BYTES = bytes(_RAW_MAGIC_NUMBER >> n & 0xff for n in range(0, 25, 8))

_PYCACHE = '__pycache__'

SOURCE_SUFFIXES = ['.py']  # _setup() adds .pyw as needed.

DEBUG_BYTECODE_SUFFIXES = ['.pyc']
OPTIMIZED_BYTECODE_SUFFIXES = ['.pyo']

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 sys.flags.optimize is non-zero, then it will be .pyo.

    If debug_override is not None, then it must be a boolean and is used in
    place of sys.flags.optimize.

    If sys.implementation.cache_tag is None then NotImplementedError is raised.

    """
    debug = not sys.flags.optimize 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('.')
    tag = sys.implementation.cache_tag
    if tag is None:
        raise NotImplementedError('sys.implementation.cache_tag is None')
    filename = ''.join([base_filename, sep, tag, suffixes[0]])
    return _path_join(head, _PYCACHE, filename)


def source_from_cache(path):
    """Given the path to a .pyc./.pyo file, return the path to its .py file.

    The .pyc/.pyo file does not need to exist; this simply returns the path to
    the .py file calculated to correspond to the .pyc/.pyo file.  If path does
    not conform to PEP 3147 format, ValueError will be raised. If
    sys.implementation.cache_tag is None then NotImplementedError is raised.

    """
    if sys.implementation.cache_tag is None:
        raise NotImplementedError('sys.implementation.cache_tag is None')
    head, pycache_filename = _path_split(path)
    head, pycache = _path_split(head)
    if pycache != _PYCACHE:
        raise ValueError('{} not bottom-level directory in '
                         '{!r}'.format(_PYCACHE, path))
    if pycache_filename.count('.') != 2:
        raise ValueError('expected only 2 dots in '
                         '{!r}'.format(pycache_filename))
    base_filename = pycache_filename.partition('.')[0]
    return _path_join(head, base_filename + SOURCE_SUFFIXES[0])


def _get_sourcefile(bytecode_path):
    """Convert a bytecode file path to a source path (if possible).

    This function exists purely for backwards-compatibility for
    PyImport_ExecCodeModuleWithFilenames() in the C API.

    """
    if len(bytecode_path) == 0:
        return None
    rest, _, extension = bytecode_path.rpartition('.')
    if not rest or extension.lower()[-3:-1] != 'py':
        return bytecode_path
    try:
        source_path = source_from_cache(bytecode_path)
    except (NotImplementedError, ValueError):
        source_path = bytecode_path[:-1]
    return source_path if _path_isfile(source_path) else bytecode_path


def _verbose_message(message, *args, verbosity=1):
    """Print the message to stderr if -v/PYTHONVERBOSE is turned on."""
    if sys.flags.verbose >= verbosity:
        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 getattr(module, '__package__', None) 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)
            # This must be done before putting the module in sys.modules
            # (otherwise an optimization shortcut in import.c becomes wrong)
            module.__initializing__ = True
            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]
        else:
            module.__initializing__ = True
        try:
            # If __package__ was not set above, __import__() will do it later.
            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("{} 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("{} is not a frozen module".format(fullname),
                              name=fullname)
        return fxn(self, fullname)
    _wrap(_requires_frozen_wrapper, fxn)
    return _requires_frozen_wrapper


def _find_module_shim(self, fullname):
    """Try to find a loader for the specified module by delegating to
    self.find_loader()."""
    # 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)
    if loader is None and len(portions):
        msg = "Not importing directory {}: missing __init__"
        _warnings.warn(msg.format(portions[0]), ImportWarning)
    return loader




# 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 _call_with_frames_removed(_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 False 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 = _call_with_frames_removed(_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 True if the frozen module is a package."""
        return _imp.is_frozen_package(fullname)


class WindowsRegistryFinder:

    """Meta path finder for modules declared in the Windows registry.
    """

    REGISTRY_KEY = (
        "Software\\Python\\PythonCore\\{sys_version}"
        "\\Modules\\{fullname}")
    REGISTRY_KEY_DEBUG = (
        "Software\\Python\\PythonCore\\{sys_version}"
        "\\Modules\\{fullname}\\Debug")
    DEBUG_BUILD = False  # Changed in _setup()

    @classmethod
    def _open_registry(cls, key):
        try:
            return _winreg.OpenKey(_winreg.HKEY_CURRENT_USER, key)
        except WindowsError:
            return _winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE, key)

    @classmethod
    def _search_registry(cls, fullname):
        if cls.DEBUG_BUILD:
            registry_key = cls.REGISTRY_KEY_DEBUG
        else:
            registry_key = cls.REGISTRY_KEY
        key = registry_key.format(fullname=fullname,
                                  sys_version=sys.version[:3])
        try:
            with cls._open_registry(key) as hkey:
                filepath = _winreg.QueryValue(hkey, "")
        except WindowsError:
            return None
        return filepath

    @classmethod
    def find_module(cls, fullname, path=None):
        """Find module named in the registry."""
        filepath = cls._search_registry(fullname)
        if filepath is None:
            return None
        try:
            _os.stat(filepath)
        except OSError:
            return None
        for loader, suffixes in _get_supported_file_loaders():
            if filepath.endswith(tuple(suffixes)):
                return loader(fullname, filepath)


class _LoaderBasics:

    """Base class of common code needed by both SourceLoader and
    SourcelessFileLoader."""

    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]
        filename_base = filename.rsplit('.', 1)[0]
        tail_name = fullname.rpartition('.')[2]
        return filename_base == '__init__' and tail_name != '__init__'

    def _bytes_from_bytecode(self, fullname, data, bytecode_path, source_stats):
        """Return the marshalled bytes from bytecode, verifying the magic
        number, timestamp and source size along the way.

        If source_stats is None then skip the timestamp check.

        """
        magic = data[:4]
        raw_timestamp = data[4:8]
        raw_size = data[8:12]
        if magic != _MAGIC_BYTES:
            msg = 'bad magic number in {!r}: {!r}'.format(fullname, magic)
            _verbose_message(msg)
            raise ImportError(msg, name=fullname, path=bytecode_path)
        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:
                    raise ImportError(
                        "bytecode is stale for {}".format(fullname),
                        name=fullname, path=bytecode_path)
        # Can't return the code object as errors from marshal loading need to
        # propagate even when source is available.
        return data[12:]

    @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:
            try:
                module.__cached__ = cache_from_source(module.__file__)
            except NotImplementedError:
                module.__cached__ = module.__file__
        else:
            module.__cached__ = module.__file__
        module.__package__ = name
        if self.is_package(name):
            module.__path__ = [_path_split(module.__file__)[0]]
        else:
            module.__package__ = module.__package__.rpartition('.')[0]
        module.__loader__ = self
        _call_with_frames_removed(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 _cache_bytecode(self, source_path, cache_path, data):
        """Optional method which writes data (bytes) to a file path (a str).

        Implementing this method allows for the writing of bytecode files.

        The source path is needed in order to correctly transfer permissions
        """
        # For backwards compatibility, we delegate to set_data()
        return self.set_data(cache_path, data)

    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 as exc:
            raise ImportError("source not available through get_data()",
                              name=fullname) from exc
        readsource = _io.BytesIO(source_bytes).readline
        try:
            encoding = tokenize.detect_encoding(readsource)
        except SyntaxError as exc:
            raise ImportError("Failed to detect encoding",
                              name=fullname) from exc
        newline_decoder = _io.IncrementalNewlineDecoder(None, True)
        try:
            return newline_decoder.decode(source_bytes.decode(encoding[0]))
        except UnicodeDecodeError as exc:
            raise ImportError("Failed to decode source file",
                              name=fullname) from exc

    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)
        source_mtime = None
        try:
            bytecode_path = cache_from_source(source_path)
        except NotImplementedError:
            bytecode_path = None
        else:
            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:
                    try:
                        bytes_data = self._bytes_from_bytecode(fullname, data,
                                                               bytecode_path,
                                                               st)
                    except (ImportError, EOFError):
                        pass
                    else:
                        _verbose_message('{} matches {}', bytecode_path,
                                        source_path)
                        found = marshal.loads(bytes_data)
                        if isinstance(found, _code_type):
                            _imp._fix_co_filename(found, source_path)
                            _verbose_message('code object from {}',
                                            bytecode_path)
                            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 = _call_with_frames_removed(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
            source_mtime is not None):
            data = bytearray(_MAGIC_BYTES)
            data.extend(_w_long(source_mtime))
            data.extend(_w_long(len(source_bytes)))
            data.extend(marshal.dumps(code_object))
            try:
                self._cache_bytecode(source_path, bytecode_path, data)
                _verbose_message('wrote {!r}', bytecode_path)
            except NotImplementedError:
                pass
        return code_object

    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.