""" Python 3 reorganized the standard library (PEP 3108). This module exposes several standard library modules to Python 2 under their new Python 3 names. It is designed to be used as follows:: from future import standard_library standard_library.install_hooks() And then these normal Py3 imports work on both Py3 and Py2:: import builtins import configparser import copyreg import queue import reprlib import socketserver import winreg # on Windows only import test.support import html, html.parser, html.entites import http, http.client, http.server import http.cookies, http.cookiejar import urllib.parse, urllib.request, urllib.response, urllib.error, urllib.robotparser import xmlrpc.client, xmlrpc.server import _thread import _dummythread import _markupbase from itertools import filterfalse, zip_longest from sys import intern (The renamed modules and functions are still available under their old names on Python 2.) To turn off the import hooks, use:: standard_library.remove_hooks() This is a cleaner alternative to this idiom (see http://docs.pythonsprints.com/python3_porting/py-porting.html):: try: import queue except ImportError: import Queue as queue Limitations ----------- We don't currently support these modules, but would like to:: import dbm import dbm.dumb import dbm.gnu import collections.abc # on Py33 import tkinter import pickle # should (optionally) bring in cPickle on Python 2 """ from __future__ import absolute_import, division, print_function import sys import logging import imp import contextlib import types import copy import os # Make a dedicated logger; leave the root logger to be configured # by the application. flog = logging.getLogger('future_stdlib') _formatter = logging.Formatter(logging.BASIC_FORMAT) _handler = logging.StreamHandler() _handler.setFormatter(_formatter) flog.addHandler(_handler) flog.setLevel(logging.WARN) from future.utils import PY2, PY3 # The modules that are defined under the same names on Py3 but with # different contents in a significant way (e.g. submodules) are: # pickle (fast one) # dbm # urllib # test # email REPLACED_MODULES = set(['test', 'urllib', 'pickle']) # add email and dbm when we support it # The following module names are not present in Python 2.x, so they cause no # potential clashes: # http # html # tkinter # xmlrpc # These modules need names from elsewhere being added to them: # subprocess: should provide getoutput and other fns from commands # module but these fns are missing: getstatus, mk2arg, # mkarg # re: needs an ASCII constant that works compatibly with Py3 # Old to new # etc: see lib2to3/fixes/fix_imports.py RENAMES = { # 'cStringIO': 'io', # there's a new io module in Python 2.6 # that provides StringIO and BytesIO # 'StringIO': 'io', # ditto # 'cPickle': 'pickle', '__builtin__': 'builtins', 'copy_reg': 'copyreg', 'Queue': 'queue', 'future.moves.socketserver': 'socketserver', 'ConfigParser': 'configparser', 'repr': 'reprlib', # 'FileDialog': 'tkinter.filedialog', # 'tkFileDialog': 'tkinter.filedialog', # 'SimpleDialog': 'tkinter.simpledialog', # 'tkSimpleDialog': 'tkinter.simpledialog', # 'tkColorChooser': 'tkinter.colorchooser', # 'tkCommonDialog': 'tkinter.commondialog', # 'Dialog': 'tkinter.dialog', # 'Tkdnd': 'tkinter.dnd', # 'tkFont': 'tkinter.font', # 'tkMessageBox': 'tkinter.messagebox', # 'ScrolledText': 'tkinter.scrolledtext', # 'Tkconstants': 'tkinter.constants', # 'Tix': 'tkinter.tix', # 'ttk': 'tkinter.ttk', # 'Tkinter': 'tkinter', '_winreg': 'winreg', 'thread': '_thread', 'dummy_thread': '_dummy_thread', # 'anydbm': 'dbm', # causes infinite import loop # 'whichdb': 'dbm', # causes infinite import loop # anydbm and whichdb are handled by fix_imports2 # 'dbhash': 'dbm.bsd', # 'dumbdbm': 'dbm.dumb', # 'dbm': 'dbm.ndbm', # 'gdbm': 'dbm.gnu', 'future.moves.xmlrpc': 'xmlrpc', # 'future.backports.email': 'email', # for use by urllib # 'DocXMLRPCServer': 'xmlrpc.server', # 'SimpleXMLRPCServer': 'xmlrpc.server', # 'httplib': 'http.client', # 'htmlentitydefs' : 'html.entities', # 'HTMLParser' : 'html.parser', # 'Cookie': 'http.cookies', # 'cookielib': 'http.cookiejar', # 'BaseHTTPServer': 'http.server', # 'SimpleHTTPServer': 'http.server', # 'CGIHTTPServer': 'http.server', # 'future.backports.test': 'test', # primarily for renaming test_support to support # 'commands': 'subprocess', # 'urlparse' : 'urllib.parse', # 'robotparser' : 'urllib.robotparser', # 'abc': 'collections.abc', # for Py33 # 'future.utils.six.moves.html': 'html', # 'future.utils.six.moves.http': 'http', 'future.moves.html': 'html', 'future.moves.http': 'http', # 'future.backports.urllib': 'urllib', # 'future.utils.six.moves.urllib': 'urllib', 'future.moves._markupbase': '_markupbase', } # It is complicated and apparently brittle to mess around with the # ``sys.modules`` cache in order to support "import urllib" meaning two # different things (Py2.7 urllib and backported Py3.3-like urllib) in different # contexts. So we require explicit imports for these modules. assert len(set(RENAMES.values()) & set(REPLACED_MODULES)) == 0 # Harmless renames that we can insert. # (New module name, new object name, old module name, old object name) MOVES = [('collections', 'UserList', 'UserList', 'UserList'), ('collections', 'UserDict', 'UserDict', 'UserDict'), ('collections', 'UserString','UserString', 'UserString'), ('itertools', 'filterfalse','itertools', 'ifilterfalse'), ('itertools', 'zip_longest','itertools', 'izip_longest'), ('sys', 'intern','__builtin__', 'intern'), # The re module has no ASCII flag in Py2, but this is the default. # Set re.ASCII to a zero constant. stat.ST_MODE just happens to be one # (and it exists on Py2.6+). ('re', 'ASCII','stat', 'ST_MODE'), ('base64', 'encodebytes','base64', 'encodestring'), ('base64', 'decodebytes','base64', 'decodestring'), ('subprocess', 'getoutput', 'commands', 'getoutput'), ('subprocess', 'getstatusoutput', 'commands', 'getstatusoutput'), ('math', 'ceil', 'future.backports.misc', 'ceil'), ('collections', 'OrderedDict', 'future.backports.misc', 'OrderedDict'), ('collections', 'Counter', 'future.backports.misc', 'Counter'), # This is no use, since "import urllib.request" etc. still fails: # ('urllib', 'error', 'future.moves.urllib', 'error'), # ('urllib', 'parse', 'future.moves.urllib', 'parse'), # ('urllib', 'request', 'future.moves.urllib', 'request'), # ('urllib', 'response', 'future.moves.urllib', 'response'), # ('urllib', 'robotparser', 'future.moves.urllib', 'robotparser'), ] # A minimal example of an import hook: # class WarnOnImport(object): # def __init__(self, *args): # self.module_names = args # # def find_module(self, fullname, path=None): # if fullname in self.module_names: # self.path = path # return self # return None # # def load_module(self, name): # if name in sys.modules: # return sys.modules[name] # module_info = imp.find_module(name, self.path) # module = imp.load_module(name, *module_info) # sys.modules[name] = module # flog.warning("Imported deprecated module %s", name) # return module class RenameImport(object): """ A class for import hooks mapping Py3 module names etc. to the Py2 equivalents. """ # Different RenameImport classes are created when importing this module from # different source files. This causes isinstance(hook, RenameImport) checks # to produce inconsistent results. We add this RENAMER attribute here so # remove_hooks() and install_hooks() can find instances of these classes # easily: RENAMER = True def __init__(self, old_to_new): ''' Pass in a dictionary-like object mapping from old names to new names. E.g. {'ConfigParser': 'configparser', 'cPickle': 'pickle'} ''' self.old_to_new = old_to_new both = set(old_to_new.keys()) & set(old_to_new.values()) assert (len(both) == 0 and len(set(old_to_new.values())) == len(old_to_new.values())), \ 'Ambiguity in renaming (handler not implemented)' self.new_to_old = dict((new, old) for (old, new) in old_to_new.items()) def find_module(self, fullname, path=None): # Handles hierarchical importing: package.module.module2 new_base_names = set([s.split('.')[0] for s in self.new_to_old]) # Before v0.12: Was: if fullname in set(self.old_to_new) | new_base_names: if fullname in new_base_names: return self return None def load_module(self, name): path = None if name in sys.modules: return sys.modules[name] elif name in self.new_to_old: # New name. Look up the corresponding old (Py2) name: oldname = self.new_to_old[name] module = self._find_and_load_module(oldname) module.__future_module__ = True else: module = self._find_and_load_module(name) # In any case, make it available under the requested (Py3) name sys.modules[name] = module return module def _find_and_load_module(self, name, path=None): """ Finds and loads it. But if there's a . in the name, handles it properly. """ bits = name.split('.') while len(bits) > 1: # Treat the first bit as a package packagename = bits.pop(0) package = self._find_and_load_module(packagename, path) try: path = package.__path__ except AttributeError: # This could be e.g. moves. flog.debug('Package {0} has no __path__.'.format(package)) if name in sys.modules: return sys.modules[name] flog.debug('What to do here?') name = bits[0] # We no longer use the fake module six.moves: # if name == 'moves': # # imp.find_module doesn't find this fake module # from future.utils.six import moves # return moves module_info = imp.find_module(name, path) return imp.load_module(name, *module_info) class hooks(object): """ Acts as a context manager. Saves the state of sys.modules and restores it after the 'with' block. Use like this: >>> from future import standard_library >>> with standard_library.hooks(): ... import http.client >>> import requests # incompatible with ``future``'s standard library hooks For this to work, http.client will be scrubbed from sys.modules after the 'with' block. That way the modules imported in the 'with' block will continue to be accessible in the current namespace but not from any imported modules (like requests). """ def __enter__(self): # flog.debug('Entering hooks context manager') self.old_sys_modules = copy.copy(sys.modules) self.hooks_were_installed = detect_hooks() self.scrubbed = scrub_py2_sys_modules() install_hooks() return self def __exit__(self, *args): # flog.debug('Exiting hooks context manager') restore_sys_modules(self.scrubbed) if not self.hooks_were_installed: remove_hooks() scrub_future_sys_modules() # Sanity check for is_py2_stdlib_module(): We aren't replacing any # builtin modules names: if PY2: assert len(set(RENAMES.values()) & set(sys.builtin_module_names)) == 0 def is_py2_stdlib_module(m): """ Tries to infer whether the module m is from the Python 2 standard library. This may not be reliable on all systems. """ if PY3: return False if not 'stdlib_path' in is_py2_stdlib_module.__dict__: stdlib_files = [contextlib.__file__, os.__file__, copy.__file__] stdlib_paths = [os.path.split(f)[0] for f in stdlib_files] if not len(set(stdlib_paths)) == 1: # This seems to happen on travis-ci.org. Very strange. We'll try to # ignore it. flog.warn('Multiple locations found for the Python standard ' 'library: %s' % stdlib_paths) # Choose the first one arbitrarily is_py2_stdlib_module.stdlib_path = stdlib_paths[0] if m.__name__ in sys.builtin_module_names: return True if hasattr(m, '__file__'): modpath = os.path.split(m.__file__) if (modpath[0].startswith(is_py2_stdlib_module.stdlib_path) and 'site-packages' not in modpath[0]): return True return False def scrub_py2_sys_modules(): """ Removes any Python 2 standard library modules from ``sys.modules`` that would interfere with Py3-style imports using import hooks. Examples are modules with the same names (like urllib or email). (Note that currently import hooks are disabled for modules like these with ambiguous names anyway ...) """ if PY3: return {} scrubbed = {} for modulename in REPLACED_MODULES & set(RENAMES.keys()): if not modulename in sys.modules: continue module = sys.modules[modulename] if is_py2_stdlib_module(module): flog.debug('Deleting (Py2) {} from sys.modules'.format(modulename)) scrubbed[modulename] = sys.modules[modulename] del sys.modules[modulename] return scrubbed def scrub_future_sys_modules(): """ On Py2 only: Removes any modules such as ``http`` and ``html.parser`` from the ``sys.modules`` cache. Such modules would confuse code such as this:: # PyChecker does something like this: try: import builtins except: PY3 = False finally: PY3 = True or this:: import urllib # We want this to pull in only the Py2 module # after scrub_future_sys_modules() has been called or this:: # Requests does this in requests/packages/urllib3/connection.py: try: # Python 3 from http.client import HTTPConnection, HTTPException except ImportError: from httplib import HTTPConnection, HTTPException This function removes items matching this spec from sys.modules:: key: new_py3_module_name value: either future.backports module or py2 module with another name """ scrubbed = {} if PY3: return {} for modulename, module in sys.modules.items(): if modulename.startswith('future'): flog.debug('Not removing %s' % modulename) continue # We don't want to remove Python 2.x urllib if this is cached. # But we do want to remove modules under their new names, e.g. # 'builtins'. # We look for builtins, configparser, urllib, email, http, etc., and # their submodules if (modulename in RENAMES.values() or any(modulename.startswith(m + '.') for m in RENAMES.values()) or 'urllib' in modulename): if module is None: # This happens for e.g. __future__ imports. Delete it. flog.debug('Deleting empty module {0} from sys.modules' .format(modulename)) del sys.modules[modulename] continue # Not all modules come from future.moves. Example: # sys.modules['builtins'] == p = os.path.join('future', 'moves', modulename.replace('.', os.sep)) # six.moves doesn't have a __file__ attribute: if (hasattr(module, '__file__') and p in module.__file__ or hasattr(module, '__future_module__')): flog.debug('Deleting (future) {0} {1} from sys.modules' .format(modulename, module)) scrubbed[modulename] = sys.modules[modulename] del sys.modules[modulename] else: flog.debug('Not deleting {0} {1} from sys.modules' .format(modulename, module)) return scrubbed class suspend_hooks(object): """ Acts as a context manager. Use like this: >>> from future import standard_library >>> standard_library.install_hooks() >>> import http.client >>> # ... >>> with standard_library.suspend_hooks(): >>> import requests # incompatible with ``future``'s standard library hooks If the hooks were disabled before the context, they are not installed when the context is left. """ def __enter__(self): self.hooks_were_installed = detect_hooks() remove_hooks() self.scrubbed = scrub_future_sys_modules() return self def __exit__(self, *args): if self.hooks_were_installed: # scrub_py2_sys_modules() # in case they interfere ... e.g. urllib install_hooks() restore_sys_modules(self.scrubbed) def restore_sys_modules(scrubbed): """ Add any previously scrubbed modules back to the sys.modules cache, but only if it's safe to do so. """ clash = set(sys.modules) & set(scrubbed) if len(clash) != 0: # If several, choose one arbitrarily to raise an exception about first = list(clash)[0] raise ImportError('future module {} clashes with Py2 module' .format(first)) sys.modules.update(scrubbed) def install_aliases(): """ Monkey-patches the standard library in Py2.6/7 to provide aliases for better Py3 compatibility. """ if PY3: return # if hasattr(install_aliases, 'run_already'): # return for (newmodname, newobjname, oldmodname, oldobjname) in MOVES: __import__(newmodname) # We look up the module in sys.modules because __import__ just returns the # top-level package: newmod = sys.modules[newmodname] newmod.__future_module__ = True __import__(oldmodname) oldmod = sys.modules[oldmodname] obj = getattr(oldmod, oldobjname) setattr(newmod, newobjname, obj) # Hack for urllib so it appears to have the same structure on Py2 as on Py3 import urllib from future.moves.urllib import request from future.moves.urllib import response from future.moves.urllib import parse from future.moves.urllib import error from future.moves.urllib import robotparser urllib.request = request urllib.response = response urllib.parse = parse urllib.error = error urllib.robotparser = robotparser sys.modules['urllib.request'] = request sys.modules['urllib.response'] = response sys.modules['urllib.parse'] = parse sys.modules['urllib.error'] = error sys.modules['urllib.robotparser'] = robotparser from future.moves import http sys.modules['http'] = http from future.moves import xmlrpc sys.modules['xmlrpc'] = xmlrpc from future.moves import html sys.modules['html'] = html # install_aliases.run_already = True def install_hooks(): """ This function installs the future.standard_library import hook into sys.meta_path. """ if PY3: return install_aliases() flog.debug('sys.meta_path was: {0}'.format(sys.meta_path)) flog.debug('Installing hooks ...') # Add it unless it's there already newhook = RenameImport(RENAMES) if not detect_hooks(): sys.meta_path.append(newhook) flog.debug('sys.meta_path is now: {0}'.format(sys.meta_path)) def enable_hooks(): """ Deprecated. Use install_hooks() instead. This will be removed by ``future`` v1.0. """ install_hooks() def remove_hooks(scrub_sys_modules=True): """ This function removes the import hook from sys.meta_path. """ if PY3: return flog.debug('Uninstalling hooks ...') # Loop backwards, so deleting items keeps the ordering: for i, hook in list(enumerate(sys.meta_path))[::-1]: if hasattr(hook, 'RENAMER'): del sys.meta_path[i] # Explicit is better than implicit. In the future the interface should # probably change so that scrubbing the import hooks requires a separate # function call. Left as is for now for backward compatibility with # v0.11.x. if scrub_sys_modules: scrub_future_sys_modules() def disable_hooks(): """ Deprecated. Use remove_hooks() instead. This will be removed by ``future`` v1.0. """ remove_hooks() def detect_hooks(): """ Returns True if the import hooks are installed, False if not. """ flog.debug('Detecting hooks ...') present = any([hasattr(hook, 'RENAMER') for hook in sys.meta_path]) if present: flog.debug('Detected.') else: flog.debug('Not detected.') return present # As of v0.12, this no longer happens implicitly: # if not PY3: # install_hooks() if not hasattr(sys, 'py2_modules'): sys.py2_modules = {} def cache_py2_modules(): """ Currently this function is unneeded, as we are not attempting to provide import hooks for modules with ambiguous names: email, urllib, pickle. """ if len(sys.py2_modules) != 0: return assert not detect_hooks() import urllib sys.py2_modules['urllib'] = urllib import email sys.py2_modules['email'] = email import pickle sys.py2_modules['pickle'] = pickle # Not all Python installations have test module. (Anaconda doesn't, for example.) # try: # import test # except ImportError: # sys.py2_modules['test'] = None # sys.py2_modules['test'] = test # import dbm # sys.py2_modules['dbm'] = dbm def import_(module_name, backport=False): """ Pass a (potentially dotted) module name of a Python 3 standard library module. This function imports the module compatibly on Py2 and Py3 and returns the top-level module. Example use: >>> http = import_('http.client') >>> http = import_('http.server') >>> urllib = import_('urllib.request') Then: >>> conn = http.client.HTTPConnection(...) >>> response = urllib.request.urlopen('http://mywebsite.com') >>> # etc. Use as follows: >>> package_name = import_(module_name) On Py3, equivalent to this: >>> import module_name On Py2, equivalent to this if backport=False: >>> from future.moves import module_name or to this if backport=True: >>> from future.backports import module_name except that it also handles dotted module names such as ``http.client`` The effect then is like this: >>> from future.backports import module >>> from future.backports.module import submodule >>> module.submodule = submodule Note that this would be a SyntaxError in Python: >>> from future.backports import http.client """ # Python 2.6 doesn't have importlib in the stdlib, so it requires # the backported ``importlib`` package from PyPI as a dependency to use # this function: import importlib if PY3: return __import__(module_name) else: # client.blah = blah # Then http.client = client # etc. if backport: prefix = 'future.backports' else: prefix = 'future.moves' parts = prefix.split('.') + module_name.split('.') modules = [] for i, part in enumerate(parts): sofar = '.'.join(parts[:i+1]) modules.append(importlib.import_module(sofar)) for i, part in reversed(list(enumerate(parts))): if i == 0: break setattr(modules[i-1], part, modules[i]) # Return the next-most top-level module after future.backports / future.moves: return modules[2] def from_import(module_name, *symbol_names, **kwargs): """ Example use: >>> HTTPConnection = from_import('http.client', 'HTTPConnection') >>> HTTPServer = from_import('http.server', 'HTTPServer') >>> urlopen, urlparse = from_import('urllib.request', 'urlopen', 'urlparse') Equivalent to this on Py3: >>> from module_name import symbol_names[0], symbol_names[1], ... and this on Py2: >>> from future.moves.module_name import symbol_names[0], ... or: >>> from future.backports.module_name import symbol_names[0], ... except that it also handles dotted module names such as ``http.client``. """ if PY3: return __import__(module_name) else: if 'backport' in kwargs and bool(kwargs['backport']): prefix = 'future.backports' else: prefix = 'future.moves' parts = prefix.split('.') + module_name.split('.') module = importlib.import_module(prefix + '.' + module_name) output = [getattr(module, name) for name in symbol_names] if len(output) == 1: return output[0] else: return output