cpython/Lib/importlib/test/source/test_abc_loader.py

877 lines
31 KiB
Python
Raw Normal View History

import importlib
from importlib import abc
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
from .. import abc as testing_abc
from .. import util
from . import util as source_util
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
import imp
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
import inspect
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
import io
import marshal
import os
import sys
import types
import unittest
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
import warnings
class SourceOnlyLoaderMock(abc.SourceLoader):
# Globals that should be defined for all modules.
source = (b"_ = '::'.join([__name__, __file__, __cached__, __package__, "
b"repr(__loader__)])")
def __init__(self, path):
self.path = path
def get_data(self, path):
assert self.path == path
return self.source
def get_filename(self, fullname):
return self.path
class SourceLoaderMock(SourceOnlyLoaderMock):
source_mtime = 1
def __init__(self, path, magic=imp.get_magic()):
super().__init__(path)
self.bytecode_path = imp.cache_from_source(self.path)
data = bytearray(magic)
data.extend(marshal._w_long(self.source_mtime))
code_object = compile(self.source, self.path, 'exec',
dont_inherit=True)
data.extend(marshal.dumps(code_object))
self.bytecode = bytes(data)
self.written = {}
def get_data(self, path):
if path == self.path:
return super().get_data(path)
elif path == self.bytecode_path:
return self.bytecode
else:
raise IOError
def path_mtime(self, path):
assert path == self.path
return self.source_mtime
def set_data(self, path, data):
self.written[path] = bytes(data)
return path == self.bytecode_path
class PyLoaderMock(abc.PyLoader):
# Globals that should be defined for all modules.
source = (b"_ = '::'.join([__name__, __file__, __package__, "
b"repr(__loader__)])")
def __init__(self, data):
"""Take a dict of 'module_name: path' pairings.
Paths should have no file extension, allowing packages to be denoted by
ending in '__init__'.
"""
self.module_paths = data
self.path_to_module = {val:key for key,val in data.items()}
def get_data(self, path):
if path not in self.path_to_module:
raise IOError
return self.source
def is_package(self, name):
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
filename = os.path.basename(self.get_filename(name))
return os.path.splitext(filename)[0] == '__init__'
def source_path(self, name):
try:
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
return self.module_paths[name]
except KeyError:
raise ImportError
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
def get_filename(self, name):
"""Silence deprecation warning."""
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
path = super().get_filename(name)
assert len(w) == 1
assert issubclass(w[0].category, PendingDeprecationWarning)
return path
class PyLoaderCompatMock(PyLoaderMock):
"""Mock that matches what is suggested to have a loader that is compatible
from Python 3.1 onwards."""
def get_filename(self, fullname):
try:
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
return self.module_paths[fullname]
except KeyError:
raise ImportError
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
def source_path(self, fullname):
try:
return self.get_filename(fullname)
except ImportError:
return None
class PyPycLoaderMock(abc.PyPycLoader, PyLoaderMock):
default_mtime = 1
def __init__(self, source, bc={}):
"""Initialize mock.
'bc' is a dict keyed on a module's name. The value is dict with
possible keys of 'path', 'mtime', 'magic', and 'bc'. Except for 'path',
each of those keys control if any part of created bytecode is to
deviate from default values.
"""
super().__init__(source)
self.module_bytecode = {}
self.path_to_bytecode = {}
self.bytecode_to_path = {}
for name, data in bc.items():
self.path_to_bytecode[data['path']] = name
self.bytecode_to_path[name] = data['path']
magic = data.get('magic', imp.get_magic())
mtime = importlib._w_long(data.get('mtime', self.default_mtime))
if 'bc' in data:
bc = data['bc']
else:
bc = self.compile_bc(name)
self.module_bytecode[name] = magic + mtime + bc
def compile_bc(self, name):
source_path = self.module_paths.get(name, '<test>') or '<test>'
code = compile(self.source, source_path, 'exec')
return marshal.dumps(code)
def source_mtime(self, name):
if name in self.module_paths:
return self.default_mtime
elif name in self.module_bytecode:
return None
else:
raise ImportError
def bytecode_path(self, name):
try:
return self.bytecode_to_path[name]
except KeyError:
if name in self.module_paths:
return None
else:
raise ImportError
def write_bytecode(self, name, bytecode):
self.module_bytecode[name] = bytecode
return True
def get_data(self, path):
if path in self.path_to_module:
return super().get_data(path)
elif path in self.path_to_bytecode:
name = self.path_to_bytecode[path]
return self.module_bytecode[name]
else:
raise IOError
def is_package(self, name):
try:
return super().is_package(name)
except TypeError:
return '__init__' in self.bytecode_to_path[name]
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
def get_code(self, name):
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
code_object = super().get_code(name)
assert len(w) == 1
assert issubclass(w[0].category, PendingDeprecationWarning)
return code_object
class PyLoaderTests(testing_abc.LoaderTests):
"""Tests for importlib.abc.PyLoader."""
mocker = PyLoaderMock
def eq_attrs(self, ob, **kwargs):
for attr, val in kwargs.items():
found = getattr(ob, attr)
self.assertEqual(found, val,
"{} attribute: {} != {}".format(attr, found, val))
def test_module(self):
name = '<module>'
path = os.path.join('', 'path', 'to', 'module')
mock = self.mocker({name: path})
with util.uncache(name):
module = mock.load_module(name)
self.assertTrue(name in sys.modules)
self.eq_attrs(module, __name__=name, __file__=path, __package__='',
__loader__=mock)
self.assertTrue(not hasattr(module, '__path__'))
return mock, name
def test_package(self):
name = '<pkg>'
path = os.path.join('path', 'to', name, '__init__')
mock = self.mocker({name: path})
with util.uncache(name):
module = mock.load_module(name)
self.assertTrue(name in sys.modules)
self.eq_attrs(module, __name__=name, __file__=path,
__path__=[os.path.dirname(path)], __package__=name,
__loader__=mock)
return mock, name
def test_lacking_parent(self):
name = 'pkg.mod'
path = os.path.join('path', 'to', 'pkg', 'mod')
mock = self.mocker({name: path})
with util.uncache(name):
module = mock.load_module(name)
self.assertIn(name, sys.modules)
self.eq_attrs(module, __name__=name, __file__=path, __package__='pkg',
__loader__=mock)
self.assertFalse(hasattr(module, '__path__'))
return mock, name
def test_module_reuse(self):
name = 'mod'
path = os.path.join('path', 'to', 'mod')
module = imp.new_module(name)
mock = self.mocker({name: path})
with util.uncache(name):
sys.modules[name] = module
loaded_module = mock.load_module(name)
self.assertTrue(loaded_module is module)
self.assertTrue(sys.modules[name] is module)
return mock, name
def test_state_after_failure(self):
name = "mod"
module = imp.new_module(name)
module.blah = None
mock = self.mocker({name: os.path.join('path', 'to', 'mod')})
mock.source = b"1/0"
with util.uncache(name):
sys.modules[name] = module
with self.assertRaises(ZeroDivisionError):
mock.load_module(name)
self.assertTrue(sys.modules[name] is module)
self.assertTrue(hasattr(module, 'blah'))
return mock
def test_unloadable(self):
name = "mod"
mock = self.mocker({name: os.path.join('path', 'to', 'mod')})
mock.source = b"1/0"
with util.uncache(name):
with self.assertRaises(ZeroDivisionError):
mock.load_module(name)
self.assertTrue(name not in sys.modules)
return mock
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
class PyLoaderCompatTests(PyLoaderTests):
"""Test that the suggested code to make a loader that is compatible from
Python 3.1 forward works."""
mocker = PyLoaderCompatMock
class PyLoaderInterfaceTests(unittest.TestCase):
"""Tests for importlib.abc.PyLoader to make sure that when source_path()
doesn't return a path everything works as expected."""
def test_no_source_path(self):
# No source path should lead to ImportError.
name = 'mod'
mock = PyLoaderMock({})
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
def test_source_path_is_None(self):
name = 'mod'
mock = PyLoaderMock({name: None})
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
def test_get_filename_with_source_path(self):
# get_filename() should return what source_path() returns.
name = 'mod'
path = os.path.join('path', 'to', 'source')
mock = PyLoaderMock({name: path})
with util.uncache(name):
self.assertEqual(mock.get_filename(name), path)
def test_get_filename_no_source_path(self):
# get_filename() should raise ImportError if source_path returns None.
name = 'mod'
mock = PyLoaderMock({name: None})
with util.uncache(name), self.assertRaises(ImportError):
mock.get_filename(name)
class PyPycLoaderTests(PyLoaderTests):
"""Tests for importlib.abc.PyPycLoader."""
mocker = PyPycLoaderMock
@source_util.writes_bytecode_files
def verify_bytecode(self, mock, name):
assert name in mock.module_paths
self.assertIn(name, mock.module_bytecode)
magic = mock.module_bytecode[name][:4]
self.assertEqual(magic, imp.get_magic())
mtime = importlib._r_long(mock.module_bytecode[name][4:8])
self.assertEqual(mtime, 1)
bc = mock.module_bytecode[name][8:]
self.assertEqual(bc, mock.compile_bc(name))
def test_module(self):
mock, name = super().test_module()
self.verify_bytecode(mock, name)
def test_package(self):
mock, name = super().test_package()
self.verify_bytecode(mock, name)
def test_lacking_parent(self):
mock, name = super().test_lacking_parent()
self.verify_bytecode(mock, name)
def test_module_reuse(self):
mock, name = super().test_module_reuse()
self.verify_bytecode(mock, name)
def test_state_after_failure(self):
super().test_state_after_failure()
def test_unloadable(self):
super().test_unloadable()
class PyPycLoaderInterfaceTests(unittest.TestCase):
"""Test for the interface of importlib.abc.PyPycLoader."""
def get_filename_check(self, src_path, bc_path, expect):
name = 'mod'
mock = PyPycLoaderMock({name: src_path}, {name: {'path': bc_path}})
with util.uncache(name):
assert mock.source_path(name) == src_path
assert mock.bytecode_path(name) == bc_path
self.assertEqual(mock.get_filename(name), expect)
def test_filename_with_source_bc(self):
# When source and bytecode paths present, return the source path.
self.get_filename_check('source_path', 'bc_path', 'source_path')
def test_filename_with_source_no_bc(self):
# With source but no bc, return source path.
self.get_filename_check('source_path', None, 'source_path')
def test_filename_with_no_source_bc(self):
# With not source but bc, return the bc path.
self.get_filename_check(None, 'bc_path', 'bc_path')
def test_filename_with_no_source_or_bc(self):
# With no source or bc, raise ImportError.
name = 'mod'
mock = PyPycLoaderMock({name: None}, {name: {'path': None}})
with util.uncache(name), self.assertRaises(ImportError):
mock.get_filename(name)
class SkipWritingBytecodeTests(unittest.TestCase):
"""Test that bytecode is properly handled based on
sys.dont_write_bytecode."""
@source_util.writes_bytecode_files
def run_test(self, dont_write_bytecode):
name = 'mod'
mock = PyPycLoaderMock({name: os.path.join('path', 'to', 'mod')})
sys.dont_write_bytecode = dont_write_bytecode
with util.uncache(name):
mock.load_module(name)
self.assertTrue((name in mock.module_bytecode) is not
dont_write_bytecode)
def test_no_bytecode_written(self):
self.run_test(True)
def test_bytecode_written(self):
self.run_test(False)
class RegeneratedBytecodeTests(unittest.TestCase):
"""Test that bytecode is regenerated as expected."""
@source_util.writes_bytecode_files
def test_different_magic(self):
# A different magic number should lead to new bytecode.
name = 'mod'
bad_magic = b'\x00\x00\x00\x00'
assert bad_magic != imp.get_magic()
mock = PyPycLoaderMock({name: os.path.join('path', 'to', 'mod')},
{name: {'path': os.path.join('path', 'to',
'mod.bytecode'),
'magic': bad_magic}})
with util.uncache(name):
mock.load_module(name)
self.assertTrue(name in mock.module_bytecode)
magic = mock.module_bytecode[name][:4]
self.assertEqual(magic, imp.get_magic())
@source_util.writes_bytecode_files
def test_old_mtime(self):
# Bytecode with an older mtime should be regenerated.
name = 'mod'
old_mtime = PyPycLoaderMock.default_mtime - 1
mock = PyPycLoaderMock({name: os.path.join('path', 'to', 'mod')},
{name: {'path': 'path/to/mod.bytecode', 'mtime': old_mtime}})
with util.uncache(name):
mock.load_module(name)
self.assertTrue(name in mock.module_bytecode)
mtime = importlib._r_long(mock.module_bytecode[name][4:8])
self.assertEqual(mtime, PyPycLoaderMock.default_mtime)
class BadBytecodeFailureTests(unittest.TestCase):
"""Test import failures when there is no source and parts of the bytecode
is bad."""
def test_bad_magic(self):
# A bad magic number should lead to an ImportError.
name = 'mod'
bad_magic = b'\x00\x00\x00\x00'
bc = {name:
{'path': os.path.join('path', 'to', 'mod'),
'magic': bad_magic}}
mock = PyPycLoaderMock({name: None}, bc)
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
def test_no_bytecode(self):
# Missing code object bytecode should lead to an EOFError.
name = 'mod'
bc = {name: {'path': os.path.join('path', 'to', 'mod'), 'bc': b''}}
mock = PyPycLoaderMock({name: None}, bc)
with util.uncache(name), self.assertRaises(EOFError):
mock.load_module(name)
def test_bad_bytecode(self):
# Malformed code object bytecode should lead to a ValueError.
name = 'mod'
bc = {name: {'path': os.path.join('path', 'to', 'mod'), 'bc': b'1234'}}
mock = PyPycLoaderMock({name: None}, bc)
with util.uncache(name), self.assertRaises(ValueError):
mock.load_module(name)
def raise_ImportError(*args, **kwargs):
raise ImportError
class MissingPathsTests(unittest.TestCase):
"""Test what happens when a source or bytecode path does not exist (either
from *_path returning None or raising ImportError)."""
def test_source_path_None(self):
# Bytecode should be used when source_path returns None, along with
# __file__ being set to the bytecode path.
name = 'mod'
bytecode_path = 'path/to/mod'
mock = PyPycLoaderMock({name: None}, {name: {'path': bytecode_path}})
with util.uncache(name):
module = mock.load_module(name)
self.assertEqual(module.__file__, bytecode_path)
# Testing for bytecode_path returning None handled by all tests where no
# bytecode initially exists.
def test_all_paths_None(self):
# If all *_path methods return None, raise ImportError.
name = 'mod'
mock = PyPycLoaderMock({name: None})
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
def test_source_path_ImportError(self):
# An ImportError from source_path should trigger an ImportError.
name = 'mod'
mock = PyPycLoaderMock({}, {name: {'path': os.path.join('path', 'to',
'mod')}})
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
def test_bytecode_path_ImportError(self):
# An ImportError from bytecode_path should trigger an ImportError.
name = 'mod'
mock = PyPycLoaderMock({name: os.path.join('path', 'to', 'mod')})
bad_meth = types.MethodType(raise_ImportError, mock)
mock.bytecode_path = bad_meth
with util.uncache(name), self.assertRaises(ImportError):
mock.load_module(name)
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
class SourceLoaderTestHarness(unittest.TestCase):
def setUp(self, *, is_package=True, **kwargs):
self.package = 'pkg'
if is_package:
self.path = os.path.join(self.package, '__init__.py')
self.name = self.package
else:
module_name = 'mod'
self.path = os.path.join(self.package, '.'.join(['mod', 'py']))
self.name = '.'.join([self.package, module_name])
self.cached = imp.cache_from_source(self.path)
self.loader = self.loader_mock(self.path, **kwargs)
def verify_module(self, module):
self.assertEqual(module.__name__, self.name)
self.assertEqual(module.__file__, self.path)
self.assertEqual(module.__cached__, self.cached)
self.assertEqual(module.__package__, self.package)
self.assertEqual(module.__loader__, self.loader)
values = module._.split('::')
self.assertEqual(values[0], self.name)
self.assertEqual(values[1], self.path)
self.assertEqual(values[2], self.cached)
self.assertEqual(values[3], self.package)
self.assertEqual(values[4], repr(self.loader))
def verify_code(self, code_object):
module = imp.new_module(self.name)
module.__file__ = self.path
module.__cached__ = self.cached
module.__package__ = self.package
module.__loader__ = self.loader
module.__path__ = []
exec(code_object, module.__dict__)
self.verify_module(module)
class SourceOnlyLoaderTests(SourceLoaderTestHarness):
"""Test importlib.abc.SourceLoader for source-only loading.
Reload testing is subsumed by the tests for
importlib.util.module_for_loader.
"""
loader_mock = SourceOnlyLoaderMock
def test_get_source(self):
# Verify the source code is returned as a string.
# If an IOError is raised by get_data then raise ImportError.
expected_source = self.loader.source.decode('utf-8')
self.assertEqual(self.loader.get_source(self.name), expected_source)
def raise_IOError(path):
raise IOError
self.loader.get_data = raise_IOError
with self.assertRaises(ImportError):
self.loader.get_source(self.name)
def test_is_package(self):
# Properly detect when loading a package.
self.setUp(is_package=True)
self.assertTrue(self.loader.is_package(self.name))
self.setUp(is_package=False)
self.assertFalse(self.loader.is_package(self.name))
def test_get_code(self):
# Verify the code object is created.
code_object = self.loader.get_code(self.name)
self.verify_code(code_object)
def test_load_module(self):
# Loading a module should set __name__, __loader__, __package__,
# __path__ (for packages), __file__, and __cached__.
# The module should also be put into sys.modules.
with util.uncache(self.name):
module = self.loader.load_module(self.name)
self.verify_module(module)
self.assertEqual(module.__path__, [os.path.dirname(self.path)])
self.assertTrue(self.name in sys.modules)
def test_package_settings(self):
# __package__ needs to be set, while __path__ is set on if the module
# is a package.
# Testing the values for a package are covered by test_load_module.
self.setUp(is_package=False)
with util.uncache(self.name):
module = self.loader.load_module(self.name)
self.verify_module(module)
self.assertTrue(not hasattr(module, '__path__'))
def test_get_source_encoding(self):
# Source is considered encoded in UTF-8 by default unless otherwise
# specified by an encoding line.
source = "_ = 'ü'"
self.loader.source = source.encode('utf-8')
returned_source = self.loader.get_source(self.name)
self.assertEqual(returned_source, source)
source = "# coding: latin-1\n_ = ü"
self.loader.source = source.encode('latin-1')
returned_source = self.loader.get_source(self.name)
self.assertEqual(returned_source, source)
@unittest.skipIf(sys.dont_write_bytecode, "sys.dont_write_bytecode is true")
class SourceLoaderBytecodeTests(SourceLoaderTestHarness):
"""Test importlib.abc.SourceLoader's use of bytecode.
Source-only testing handled by SourceOnlyLoaderTests.
"""
loader_mock = SourceLoaderMock
def verify_code(self, code_object, *, bytecode_written=False):
super().verify_code(code_object)
if bytecode_written:
self.assertIn(self.cached, self.loader.written)
data = bytearray(imp.get_magic())
data.extend(marshal._w_long(self.loader.source_mtime))
data.extend(marshal.dumps(code_object))
self.assertEqual(self.loader.written[self.cached], bytes(data))
def test_code_with_everything(self):
# When everything should work.
code_object = self.loader.get_code(self.name)
self.verify_code(code_object)
def test_no_bytecode(self):
# If no bytecode exists then move on to the source.
self.loader.bytecode_path = "<does not exist>"
# Sanity check
with self.assertRaises(IOError):
bytecode_path = imp.cache_from_source(self.path)
self.loader.get_data(bytecode_path)
code_object = self.loader.get_code(self.name)
self.verify_code(code_object, bytecode_written=True)
def test_code_bad_timestamp(self):
# Bytecode is only used when the timestamp matches the source EXACTLY.
for source_mtime in (0, 2):
assert source_mtime != self.loader.source_mtime
original = self.loader.source_mtime
self.loader.source_mtime = source_mtime
# If bytecode is used then EOFError would be raised by marshal.
self.loader.bytecode = self.loader.bytecode[8:]
code_object = self.loader.get_code(self.name)
self.verify_code(code_object, bytecode_written=True)
self.loader.source_mtime = original
def test_code_bad_magic(self):
# Skip over bytecode with a bad magic number.
self.setUp(magic=b'0000')
# If bytecode is used then EOFError would be raised by marshal.
self.loader.bytecode = self.loader.bytecode[8:]
code_object = self.loader.get_code(self.name)
self.verify_code(code_object, bytecode_written=True)
def test_dont_write_bytecode(self):
# Bytecode is not written if sys.dont_write_bytecode is true.
# Can assume it is false already thanks to the skipIf class decorator.
try:
sys.dont_write_bytecode = True
self.loader.bytecode_path = "<does not exist>"
code_object = self.loader.get_code(self.name)
self.assertNotIn(self.cached, self.loader.written)
finally:
sys.dont_write_bytecode = False
def test_no_set_data(self):
# If set_data is not defined, one can still read bytecode.
self.setUp(magic=b'0000')
original_set_data = self.loader.__class__.set_data
try:
del self.loader.__class__.set_data
code_object = self.loader.get_code(self.name)
self.verify_code(code_object)
finally:
self.loader.__class__.set_data = original_set_data
def test_set_data_raises_exceptions(self):
# Raising NotImplementedError or IOError is okay for set_data.
def raise_exception(exc):
def closure(*args, **kwargs):
raise exc
return closure
self.setUp(magic=b'0000')
self.loader.set_data = raise_exception(NotImplementedError)
code_object = self.loader.get_code(self.name)
self.verify_code(code_object)
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
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
class SourceLoaderGetSourceTests(unittest.TestCase):
"""Tests for importlib.abc.SourceLoader.get_source()."""
def test_default_encoding(self):
# Should have no problems with UTF-8 text.
name = 'mod'
mock = SourceOnlyLoaderMock('mod.file')
source = 'x = "ü"'
mock.source = source.encode('utf-8')
returned_source = mock.get_source(name)
self.assertEqual(returned_source, source)
def test_decoded_source(self):
# Decoding should work.
name = 'mod'
mock = SourceOnlyLoaderMock("mod.file")
source = "# coding: Latin-1\nx='ü'"
assert source.encode('latin-1') != source.encode('utf-8')
mock.source = source.encode('latin-1')
returned_source = mock.get_source(name)
self.assertEqual(returned_source, source)
def test_universal_newlines(self):
# PEP 302 says universal newlines should be used.
name = 'mod'
mock = SourceOnlyLoaderMock('mod.file')
source = "x = 42\r\ny = -13\r\n"
mock.source = source.encode('utf-8')
expect = io.IncrementalNewlineDecoder(None, True).decode(source)
self.assertEqual(mock.get_source(name), expect)
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
class AbstractMethodImplTests(unittest.TestCase):
"""Test the concrete abstractmethod implementations."""
class Loader(abc.Loader):
def load_module(self, fullname):
super().load_module(fullname)
class Finder(abc.Finder):
def find_module(self, _):
super().find_module(_)
class ResourceLoader(Loader, abc.ResourceLoader):
def get_data(self, _):
super().get_data(_)
class InspectLoader(Loader, abc.InspectLoader):
def is_package(self, _):
super().is_package(_)
def get_code(self, _):
super().get_code(_)
def get_source(self, _):
super().get_source(_)
class ExecutionLoader(InspectLoader, abc.ExecutionLoader):
def get_filename(self, _):
super().get_filename(_)
class SourceLoader(ResourceLoader, ExecutionLoader, abc.SourceLoader):
pass
class PyLoader(ResourceLoader, InspectLoader, abc.PyLoader):
def source_path(self, _):
super().source_path(_)
class PyPycLoader(PyLoader, abc.PyPycLoader):
def bytecode_path(self, _):
super().bytecode_path(_)
def source_mtime(self, _):
super().source_mtime(_)
def write_bytecode(self, _, _2):
super().write_bytecode(_, _2)
def raises_NotImplementedError(self, ins, *args):
for method_name in args:
method = getattr(ins, method_name)
arg_count = len(inspect.getfullargspec(method)[0]) - 1
args = [''] * arg_count
try:
method(*args)
except NotImplementedError:
pass
else:
msg = "{}.{} did not raise NotImplementedError"
self.fail(msg.format(ins.__class__.__name__, method_name))
def test_Loader(self):
self.raises_NotImplementedError(self.Loader(), 'load_module')
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
# XXX misplaced; should be somewhere else
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
def test_Finder(self):
self.raises_NotImplementedError(self.Finder(), 'find_module')
def test_ResourceLoader(self):
self.raises_NotImplementedError(self.ResourceLoader(), 'load_module',
'get_data')
def test_InspectLoader(self):
self.raises_NotImplementedError(self.InspectLoader(), 'load_module',
'is_package', 'get_code', 'get_source')
def test_ExecutionLoader(self):
self.raises_NotImplementedError(self.ExecutionLoader(), 'load_module',
'is_package', 'get_code', 'get_source',
'get_filename')
def test_SourceLoader(self):
ins = self.SourceLoader()
# Required abstractmethods.
self.raises_NotImplementedError(ins, 'get_filename', 'get_data')
# Optional abstractmethods.
self.raises_NotImplementedError(ins,'path_mtime', 'set_data')
def test_PyLoader(self):
self.raises_NotImplementedError(self.PyLoader(), 'source_path',
'get_data', 'is_package')
def test_PyPycLoader(self):
self.raises_NotImplementedError(self.PyPycLoader(), 'source_path',
'source_mtime', 'bytecode_path',
'write_bytecode')
def test_main():
from test.support import run_unittest
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
run_unittest(PyLoaderTests, PyLoaderCompatTests,
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
PyLoaderInterfaceTests,
PyPycLoaderTests, PyPycLoaderInterfaceTests,
SkipWritingBytecodeTests, RegeneratedBytecodeTests,
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
BadBytecodeFailureTests, MissingPathsTests,
SourceOnlyLoaderTests,
SourceLoaderBytecodeTests,
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
SourceLoaderGetSourceTests,
Implement importlib.abc.SourceLoader and deprecate PyLoader and PyPycLoader. SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only wants to use source, then they need to only implement get_data and get_filename. To also use bytecode -- sourceless loading is not supported -- then two abstract methods -- path_mtime and set_data -- need to be implemented. Compared to PyLoader and PyPycLoader, there are less abstract methods introduced and bytecode files become an optimization controlled by the ABC and hidden from the user (this need came about as PEP 3147 showed that not treating bytecode as an optimization can cause problems for compatibility). PyLoader is deprecated in favor of SourceLoader. To be compatible from Python 3.1 onwards, a subclass need only use simple methods for source_path and is_package. Otherwise conditional subclassing based on whether Python 3.1 or Python 3.2 is being is the only change. The documentation and docstring for PyLoader explain what is exactly needed. PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147 shifted bytecode path details so much, there is no foolproof way to provide backwards-compatibility with SourceLoader. Because of this the class is simply deprecated and users should move to SourceLoader (and optionally PyLoader for Python 3.1). This does lead to a loss of support for sourceless loading unfortunately. At some point before Python 3.2 is released, SourceLoader will be moved over to importlib._bootstrap so that the core code of importlib relies on the new code instead of the old PyPycLoader code. This commit is being done now so that there is no issue in having the API in Python 3.1a1.
2010-06-27 20:57:46 -03:00
AbstractMethodImplTests)
if __name__ == '__main__':
test_main()