注:本文的作者是 Ed Schofield。原文的地址为: Cheat Sheet: Writing Python 2-3 compatible code

该笔记向你展示了编写不会过时的兼容 Python 2 和 Python 3 的代码风格。

它是 Ed Schofield 在 PyCon AU 2014 的演讲,“ Writing 2/3 compatible code ”。

最低版本:

  • Python 2: 2.6+
  • Python 3: 3.3+

安装

一些下面的 imports 适用于使用 pip 安装在 PyPI 上安装:

import future        # pip install future
import past          # pip install past
import six           # pip install six

以下的脚本也适用于 pip 安装:

futurize             # pip install futurize
pasteurize           # pip install pasteurize

查看 http://python-future.org 和 https://pythonhosted.org/six/ 获取更多消息。

基本语法差异

print

# Python 2 only:
print \'Hello\'
# Python 2 and 3:
print(\'Hello\')

为了打印出多个 strings。 import print_function 来防止 Py2 把它解释成一个元组。

# Python 2 only:
print \'Hello\', \'Guido\'
# Python 2 and 3:
from __future__ import print_function    # (at top of module)

print(\'Hello\', \'Guido\')
# Python 2 only:
print >> sys.stderr, \'Hello\'
# Python 2 and 3:
from __future__ import print_function

print(\'Hello\', file=sys.stderr)
# Python 2 only:
print \'Hello\',
# Python 2 and 3:
from __future__ import print_function

print(\'Hello\', end=\'\')

抛出异常

# Python 2 only:
raise ValueError, \"dodgy value\"
# Python 2 and 3:
raise ValueError(\"dodgy value\")

使用 traceback 抛出异常:

# Python 2 only:
traceback = sys.exc_info()[2]
raise ValueError, \"dodgy value\", traceback
# Python 3 only:
raise ValueError(\"dodgy value\").with_traceback()
# Python 2 and 3: option 1
from six import reraise as raise_
# or
from future.utils import raise_

traceback = sys.exc_info()[2]
raise_(ValueError, \"dodgy value\", traceback)
# Python 2 and 3: option 2
from future.utils import raise_with_traceback

raise_with_traceback(ValueError(\"dodgy value\"))

异常链 (PEP 3134):

# Setup:
class DatabaseError(Exception):
    pass
# Python 3 only
class FileDatabase:
    def __init__(self, filename):
        try:
            self.file = open(filename)
        except IOError as exc:
            raise DatabaseError(\'failed to open\') from exc
# Python 2 and 3:
from future.utils import raise_from

class FileDatabase:
    def __init__(self, filename):
        try:
            self.file = open(filename)
        except IOError as exc:
            raise_from(DatabaseError(\'failed to open\'), exc)
# Testing the above:
try:
    fd = FileDatabase(\'non_existent_file.txt\')
except Exception as e:
    assert isinstance(e.__cause__, IOError)    # FileNotFoundError on Py3.3+ inherits from IOError

捕获异常

# Python 2 only:
try:
    ...
except ValueError, e:
    ...
# Python 2 and 3:
try:
    ...
except ValueError as e:
    ...

除法

整除(rounding down):

# Python 2 only:
assert 2 / 3 == 0
# Python 2 and 3:
assert 2 // 3 == 0

“True division” (float division):

# Python 3 only:
assert 3 / 2 == 1.5
# Python 2 and 3:
from __future__ import division    # (at top of module)

assert 3 / 2 == 1.5

“Old division” (i.e. compatible with Py2 behaviour):

# Python 2 only:
a = b / c            # with any types
# Python 2 and 3:
from past.utils import old_div

a = old_div(b, c)    # always same as / on Py2

长整数

短整数在 Python 3 中已经去除了,并且 long 已经变成了 int(没有 Lrepr 后面)

# Python 2 only
k = 9223372036854775808L

# Python 2 and 3:
k = 9223372036854775808
# Python 2 only
bigint = 1L

# Python 2 and 3
from future.builtins import int
bigint = int(1)

为了测试一个值是否是整数(任何类型):

# Python 2 only:
if isinstance(x, (int, long)):
    ...

# Python 3 only:
if isinstance(x, int):
    ...

# Python 2 and 3: option 1
from future.builtins import int    # subclass of long on Py2

if isinstance(x, int):             # matches both int and long on Py2
    ...

# Python 2 and 3: option 2
from past.builtins import long

if isinstance(x, (int, long)):
    ...

八进制常量

0644     # Python 2 only
0o644    # Python 2 and 3

Backtick repr

`x`      # Python 2 only
repr(x)  # Python 2 and 3

元类

class BaseForm(object):
    pass

class FormType(type):
    pass
# Python 2 only:
class Form(BaseForm):
    __metaclass__ = FormType
    pass
# Python 3 only:
class Form(BaseForm, metaclass=FormType):
    pass
# Python 2 and 3:
from six import with_metaclass
# or
from future.utils import with_metaclass

class Form(with_metaclass(FormType, BaseForm)):
    pass

Strings 和 bytes

Unicode(text) 字符字面量

如果你更新一个已经存在的 Python 2 的代码库,为所有的字符字面量使用 u 做前缀作为 unicode 的标记可能会更好。

# Python 2 only
s1 = \'The Zen of Python\'
s2 = u\'きたないのよりきれいな方がいい\\n\'

# Python 2 and 3
s1 = u\'The Zen of Python\'
s2 = u\'きたないのよりきれいな方がいい\\n\'

futurizepython-modernize 工具没有提供一个选项自动完成这个。

如果你为了个项目编写新的代码库,你可以用一个 unicode 字符模块标记所有的字符字面量。

# Python 2 and 3
from __future__ import unicode_literals    # at top of module

s1 = \'The Zen of Python\'
s2 = \'きたないのよりきれいな方がいい\\n\'

查看 http://python-future.org/unicode_literals.html 这里获取更多的关于讨论使用哪种风格的讨论。

字节-字符 字面量

# Python 2 only
s = \'This must be a byte-string\'

# Python 2 and 3
s = b\'This must be a byte-string\'

To loop over a byte-string with possible high-bit characters, obtaining each character as a byte-string of length 1:

# Python 2 only:
for bytechar in \'byte-string with high-bit chars like \\xf9\':
    ...

# Python 3 only:
for myint in b\'byte-string with high-bit chars like \\xf9\':
    bytechar = bytes([myint])

# Python 2 and 3:
from future.builtins import bytes
for myint in bytes(b\'byte-string with high-bit chars like \\xf9\'):
    bytechar = bytes([myint])

chr().encode(\'latin-1\') 其中的任何一个都可以用于把一个 int 转换成一个 1-char byte string

# Python 3 only:
for myint in b\'byte-string with high-bit chars like \\xf9\':
    char = chr(myint)    # returns a unicode string
    bytechar = char.encode(\'latin-1\')

# Python 2 and 3:
from future.builtins import bytes, chr
for myint in bytes(b\'byte-string with high-bit chars like \\xf9\'):
    char = chr(myint)    # returns a unicode string
    bytechar = char.encode(\'latin-1\')    # forces returning a byte str

basestring

# Python 2 only:
a = u\'abc\'
b = \'def\'
assert (isinstance(a, basestring) and isinstance(b, basestring))

# Python 2 and 3: alternative 1
from past.builtins import basestring    # pip install future

a = u\'abc\'
b = b\'def\'
assert (isinstance(a, basestring) and isinstance(b, basestring))
# Python 2 and 3: alternative 2: refactor the code to avoid considering
# byte-strings as strings.

from future.builtins import str
a = u\'abc\'
b = b\'def\'
c = b.decode()
assert isinstance(a, str) and isinstance(c, str)
# ...

unicode

# Python 2 only:
templates = [u\"blog/blog_post_detail_%s.html\" % unicode(slug)]
# Python 2 and 3: alternative 1
from future.builtins import str
templates = [u\"blog/blog_post_detail_%s.html\" % str(slug)]
# Python 2 and 3: alternative 2
from future.builtins import str as text
templates = [u\"blog/blog_post_detail_%s.html\" % text(slug)]

StringIO

# Python 2 only:
from StringIO import StringIO
# or:
from cStringIO import StringIO

# Python 2 and 3:
from io import BytesIO     # for handling byte strings
from io import StringIO    # for handling unicode strings

Imports relative to a package

假设包的结构是这样的:

mypackage/
    __init__.py
    submodule1.py
    submodule2.py

submodule1.py 的代码如下:

# Python 2 only:
import submodule2
# Python 2 and 3:
from . import submodule2
# Python 2 and 3:
# To make Py2 code safer (more like Py3) by preventing
# implicit relative imports, you can also add this to the top:
from __future__ import absolute_import

字典

heights = {\'Fred\': 175, \'Anne\': 166, \'Joe\': 192}

Iterating through dict keys/values/items

迭代字典的键:

# Python 2 only:
for key in heights.iterkeys():
    ...
# Python 2 and 3:
for key in heights:
    ...

迭代字典的值:

# Python 2 only:
for value in heights.itervalues():
    ...
# Idiomatic Python 3
for value in heights.values():    # extra memory overhead on Py2
    ...
# Python 2 and 3: option 1
from future.builtins import dict

heights = dict(Fred=175, Anne=166, Joe=192)
for key in heights.values():    # efficient on Py2 and Py3
    ...
# Python 2 and 3: option 2
from future.builtins import itervalues
# or
from six import itervalues

for key in itervalues(heights):
    ...

迭代字典元素:

# Python 2 only:
for (key, value) in heights.iteritems():
    ...
# Python 2 and 3: option 1
for (key, value) in heights.items():    # inefficient on Py2
    ...
# Python 2 and 3: option 2
from future.builtins import iteritems
# or
from six import iteritems

for (key, value) in iteritems(heights):
    ...

字典的键/值/元素 作为一个列表

字典的键作为一个列表:

# Python 2 only:
keylist = heights.keys()
assert isinstance(keylist, list)
# Python 2 and 3:
keylist = list(heights)
assert isinstance(keylist, list)

字典的值作为一个列表:

# Python 2 only:
heights = {\'Fred\': 175, \'Anne\': 166, \'Joe\': 192}
valuelist = heights.values()
assert isinstance(valuelist, list)
# Python 2 and 3: option 1
valuelist = list(heights.values())    # inefficient on Py2
# Python 2 and 3: option 2
from future.builtins import dict

heights = dict(Fred=175, Anne=166, Joe=192)
valuelist = list(heights.values())
# Python 2 and 3: option 3
from future.utils import listvalues

valuelist = listvalues(heights)
# Python 2 and 3: option 4
from future.utils import itervalues
# or
from six import itervalues

valuelist = list(itervalues(heights))

字典元素作为一个列表:

# Python 2 and 3: option 1
itemlist = list(heights.items())    # inefficient on Py2
# Python 2 and 3: option 2
from future.utils import listitems

itemlist = listitems(heights)
# Python 2 and 3: option 3
from future.utils import iteritems
# or
from six import iteritems

itemlist = list(iteritems(heights))

自定义类的行为

自定义迭代器

# Python 2 only
class Upper(object):
    def __init__(self, iterable):
        self._iter = iter(iterable)
    def next(self):          # Py2-style
        return self._iter.next().upper()
    def __iter__(self):
        return self

itr = Upper(\'hello\')
assert itr.next() == \'H\'     # Py2-style
assert list(itr) == list(\'ELLO\')
# Python 2 and 3: option 1
from future.builtins import object

class Upper(object):
    def __init__(self, iterable):
        self._iter = iter(iterable)
    def __next__(self):      # Py3-style iterator interface
        return next(self._iter).upper()  # builtin next() function calls
    def __iter__(self):
        return self

itr = Upper(\'hello\')
assert next(itr) == \'H\'      # compatible style
assert list(itr) == list(\'ELLO\')
# Python 2 and 3: option 2
from future.utils import implements_iterator

@implements_iterator
class Upper(object):
    def __init__(self, iterable):
        self._iter = iter(iterable)
    def __next__(self):                  # Py3-style iterator interface
        return next(self._iter).upper()  # builtin next() function calls
    def __iter__(self):
        return self

itr = Upper(\'hello\')
assert next(itr) == \'H\'
assert list(itr) == list(\'ELLO\')

自定义 __str__ 方法

# Python 2 only:
class MyClass(object):
    def __unicode__(self):
        return \'Unicode string: \\u5b54\\u5b50\'
    def __str__(self):
        return unicode(self).encode(\'utf-8\')

a = MyClass()
print(a)    # prints encoded string
# Python 2 and 3:
from future.utils import python_2_unicode_compatible

@python_2_unicode_compatible
class MyClass(object):
    def __str__(self):
        return u\'Unicode string: \\u5b54\\u5b50\'

a = MyClass()
print(a)    # prints string encoded as utf-8 on Py2
Unicode string: 孔子

自定义 __nonzero__ 对比 __bool__ 方法

# Python 2 only:
class AllOrNothing(object):
    def __init__(self, l):
        self.l = l
    def __nonzero__(self):
        return all(self.l)

container = AllOrNothing([0, 100, 200])
assert not bool(container)
# Python 2 and 3:
from future.builtins import object

class AllOrNothing(object):
    def __init__(self, l):
        self.l = l
    def __bool__(self):
        return all(self.l)

container = AllOrNothing([0, 100, 200])
assert not bool(container)

列表对迭代器

xrange

# Python 2 only:
for i in xrange(10**8):
    ...
# Python 2 and 3: forward-compatible
from future.builtins import range
for i in range(10**8):
    ...
# Python 2 and 3: backward-compatible
from past.builtins import xrange
for i in xrange(10**8):
    ...

range

# Python 2 only
mylist = range(5)
assert mylist == [0, 1, 2, 3, 4]
# Python 2 and 3: forward-compatible: option 1
mylist = list(range(5))            # copies memory on Py2
assert mylist == [0, 1, 2, 3, 4]
# Python 2 and 3: forward-compatible: option 2
from future.builtins import range

mylist = list(range(5))
assert mylist == [0, 1, 2, 3, 4]
# Python 2 and 3: option 3
from future.utils import lrange

mylist = lrange(5)
assert mylist == [0, 1, 2, 3, 4]
# Python 2 and 3: backward compatible
from past.builtins import range

mylist = range(5)
assert mylist == [0, 1, 2, 3, 4]

map

# Python 2 only:
mynewlist = map(f, myoldlist)
assert mynewlist == [f(x) for x in myoldlist]
# Python 2 and 3: option 1
# Idiomatic Py3, but inefficient on Py2
mynewlist = list(map(f, myoldlist))
assert mynewlist == [f(x) for x in myoldlist]
# Python 2 and 3: option 2
from future.builtins import map

mynewlist = list(map(f, myoldlist))
assert mynewlist == [f(x) for x in myoldlist]
# Python 2 and 3: option 3
try:
    import itertools.imap as map
except ImportError:
    pass

mynewlist = list(map(f, myoldlist))    # inefficient on Py2
assert mynewlist == [f(x) for x in myoldlist]
# Python 2 and 3: option 4
from future.utils import lmap

mynewlist = lmap(f, myoldlist)
assert mynewlist == [f(x) for x in myoldlist]
# Python 2 and 3: option 5
from past.builtins import map

mynewlist = map(f, myoldlist)
assert mynewlist == [f(x) for x in myoldlist]

imap

# Python 2 only:
from itertools import imap

myiter = imap(func, myoldlist)
assert isinstance(myiter, iter)
# Python 3 only:
myiter = map(func, myoldlist)
assert isinstance(myiter, iter)
# Python 2 and 3: option 1
from future.builtins import map

myiter = map(func, myoldlist)
assert isinstance(myiter, iter)
# Python 2 and 3: option 2
try:
    import itertools.imap as map
except ImportError:
    pass

myiter = map(func, myoldlist)
assert isinstance(myiter, iter)

zip, izip

As above with zip and itertools.izip.

filter, ifilter

As above with filter and itertools.ifilter too

其他内建函数

File IO with open()

# Python 2 only
f = open(\'myfile.txt\')
data = f.read()              # as a byte string
text = data.decode(\'utf-8\')

# Python 2 and 3: alternative 1
from io import open
f = open(\'myfile.txt\', \'rb\')
data = f.read()              # as bytes
text = data.decode(\'utf-8\')  # unicode, not bytes

# Python 2 and 3: alternative 2
from io import open
f = open(\'myfile.txt\', encoding=\'utf-8\')
text = f.read()    # unicode, not bytes

reduce()

# Python 2 only:
assert reduce(lambda x, y: x+y, [1, 2, 3, 4, 5]) == 1+2+3+4+5
# Python 2 and 3:
from functools import reduce

assert reduce(lambda x, y: x+y, [1, 2, 3, 4, 5]) == 1+2+3+4+5

raw_input()

# Python 2 only:
name = raw_input(\'What is your name? \')
assert isinstance(name, str)    # native str
# Python 2 and 3:
from future.builtins import input

name = input(\'What is your name? \')
assert isinstance(name, str)    # native str on Py2 and Py3

input()

# Python 2 only:
input(\"Type something safe please: \")
# Python 2 and 3
from future.builtins import input
eval(input(\"Type something safe please: \"))

警告:使用上面任何一个都是不安全的

file()

# Python 2 only:
f = file(pathname)
# Python 2 and 3:
f = open(pathname)

# But preferably, use this:
from io import open
f = open(pathname, \'rb\')   # if f.read() should return bytes
# or
f = open(pathname, \'rt\')   # if f.read() should return unicode text

execfile()

# Python 2 only:
execfile(\'myfile.py\')
# Python 2 and 3: alternative 1
from past.builtins import execfile

execfile(\'myfile.py\')
# Python 2 and 3: alternative 2
exec(compile(open(\'myfile.py\').read()))

# This can sometimes cause this:
#     SyntaxError: function ... uses import * and bare exec ...
# See https://github.com/PythonCharmers/python-future/issues/37

unichr()

# Python 2 only:
assert unichr(8364) == \'€\'
# Python 3 only:
assert chr(8364) == \'€\'
# Python 2 and 3:
from future.builtins import chr
assert chr(8364) == \'€\'

intern()

# Python 2 only:
intern(\'mystring\')
# Python 3 only:
from sys import intern
intern(\'mystring\')
# Python 2 and 3: alternative 1
from past.builtins import intern
intern(\'mystring\')
# Python 2 and 3: alternative 2
try:
    from sys import intern
except ImportError:
    pass
intern(\'mystring\')

apply()

args = (\'a\', \'b\')
kwargs = {\'kwarg1\': True}
# Python 2 only:
apply(f, args, kwargs)
# Python 2 and 3: alternative 1
f(*args, **kwargs)
# Python 2 and 3: alternative 2
from past.builtins import apply
apply(f, args, kwargs)

chr()

# Python 2 only:
assert chr(64) == b\'@\'
assert chr(200) == b\'\\xc8\'
# Python 3 only: option 1
assert chr(64).encode(\'latin-1\') == b\'@\'
assert chr(0xc8).encode(\'latin-1\') == b\'\\xc8\'
# Python 2 and 3: option 1
from future.builtins import chr

assert chr(64).encode(\'latin-1\') == b\'@\'
assert chr(0xc8).encode(\'latin-1\') == b\'\\xc8\'
# Python 3 only: option 2
assert bytes([64]) == b\'@\'
assert bytes([0xc8]) == b\'\\xc8\'
# Python 2 and 3: option 2
from future.builtins import bytes

assert bytes([64]) == b\'@\'
assert bytes([0xc8]) == b\'\\xc8\'

cmp()

# Python 2 only:
assert cmp(\'a\', \'b\') < 0 and cmp(\'b\', \'a\') > 0 and cmp(\'c\', \'c\') == 0
# Python 2 and 3: alternative 1
from past.builtins import cmp
assert cmp(\'a\', \'b\') < 0 and cmp(\'b\', \'a\') > 0 and cmp(\'c\', \'c\') == 0
# Python 2 and 3: alternative 2
cmp = lambda(x, y): (x > y) - (x < y)
assert cmp(\'a\', \'b\') < 0 and cmp(\'b\', \'a\') > 0 and cmp(\'c\', \'c\') == 0

reload()

# Python 2 only:
reload(mymodule)
# Python 2 and 3
from imp import reload
reload(mymodule)

标准库

StringIO 模块

# Python 2 only
from StringIO import StringIO
from cStringIO import StringIO
# Python 2 and 3
from io import BytesIO
# and refactor StringIO() calls to BytesIO() if passing byte-strings

http 模块

# Python 2 only:
import httplib
import Cookie
import cookielib
import BaseHTTPServer
import SimpleHTTPServer
import CGIHttpServer

# Python 2 and 3:
from future.standard_library import hooks
with hooks():
    import http.client
    import http.cookies
    import http.cookiejar
    import http.server

urllib 模块

这个使用 urllib 的示例,相同的模式也适用于其他的已经移除的模块(一个完整的列表 http://python-future.org/standard_library_imports.html#list-of-standard-library-modules )

# Python 2 only:
from urlparse import urlparse
from urllib import urlencode
# Python 3 only:
from urllib.parse import urlparse, urlencode
# Python 2 and 3: alternative 1
from future.standard_library import hooks

with hooks():
    from urllib.parse import urlparse, urlencode
# Python 2 and 3: alternative 2
from future.moves.urllib.parse import urlparse, urlencode
# or
from six.moves.urllib.parse import urlparse, urlencode
# Python 2 and 3: alternative 3
try:
    from urllib.parse import urlparse, urlencode
except ImportError:
    from urlparse import urlparse
    from urllib import urlencode