編寫兼容 Python 2 和 Python 3 的代碼

注：本文的做者是 Ed Schofield。原文的地址爲： Cheat Sheet: Writing Python 2-3 compatible code

該筆記向你展現了編寫不會過期的兼容 Python 2 和 Python 3 的代碼風格。

它是 Ed Schofield 在 PyCon AU 2014 的演講，「 Writing 2/3 compatible code 」。

最低版本：

• Python 2: 2.6+
• Python 3: 3.3+

安裝

一些下面的 imports 適用於使用 pip 安裝在 PyPI 上安裝：

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

如下的腳本也適用於 pip 安裝：

futurize             # pip install futurize
pasteurize           # pip install pasteurize

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

基本語法差別

print：

# Python 2 only:
print 'Hello'

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

爲了打印出多個 strings。 import print_function 來防止 Py2 把它解釋成一個元組。

# 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='')

拋出異常

# Python 2 only:
raise ValueError, "dodgy value"

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

使用 traceback 拋出異常：

# 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"))

異常鏈 (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

捕獲異常：

# 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

長整數

短整數在 Python 3 中已經去除了，而且 long 已經變成了 int（沒有 L 在 repr 後面）

# 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)

爲了測試一個值是不是整數（任何類型）：

# 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)):
    ...

八進制常量

0644     # Python 2 only

0o644    # Python 2 and 3

Backtick repr

`x`      # Python 2 only

repr(x)  # Python 2 and 3

元類

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

若是你更新一個已經存在的 Python 2 的代碼庫，爲全部的字符字面量使用 u 作前綴做爲 unicode 的標記可能會更好。

# 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'

futurize 和 python-modernize 工具沒有提供一個選項自動完成這個。

若是你爲了個項目編寫新的代碼庫，你能夠用一個 unicode 字符模塊標記全部的字符字面量。

# 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 這裏獲取更多的關於討論使用哪一種風格的討論。

字節-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() 和 .encode('latin-1') 其中的任何一個均可以用於把一個 int 轉換成一個 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

假設包的結構是這樣的：

mypackage/
    __init__.py
    submodule1.py
    submodule2.py

submodule1.py 的代碼以下：

# 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

迭代字典的鍵：

# 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):
    ...

字典的鍵/值/元素 做爲一個列表

字典的鍵做爲一個列表：

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

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

字典的值做爲一個列表：

# 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))

字典元素做爲一個列表：

# 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))

自定義類的行爲

自定義迭代器：

# 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')

自定義 __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

自定義 __nonzero__ 對比 __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)

列表對迭代器

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

其餘內建函數

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：使用上面任何一個都是不安全的

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)

標準庫

StringIO 模塊

# 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 模塊

# 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 模塊

這個使用 urllib 的示例，相同的模式也適用於其餘的已經移除的模塊（一個完整的列表 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