建议的解决方案 here 对我来说仍然很好(下面是代码的修改版本,仍然是在python 2中,而不是python 3中,还有一个用法示例):

#!/usr/bin/python
# -*- coding: utf-8 -*-

import codecs, logging, sys
logging.basicConfig(level=logging.INFO)
bomdict = {
    codecs.BOM_UTF8 : 'UTF8',
    codecs.BOM_UTF16_BE : 'UTF-16BE',
    codecs.BOM_UTF16_LE : 'UTF-16LE' }

def read_unicode(filename):
  the_text = open(filename, 'r').read()
  for bom, encoding in bomdict.items():
      if the_text.startswith(bom):
          logging.info('BOM found, using %s', encoding)
          the_text = the_text[len(bom):]
          break
  else:
      logging.info('No BOM, using utf8')
      encoding = 'UTF8'
  return the_text.decode(encoding)

f = open('x.txt', 'wb')
f.write(codecs.BOM_UTF16_LE)
f.write(u'zeé fóo!'.encode('UTF-16LE'))
f.close()

print read_unicode('x.txt')

以下是文件.open()的部分工作替换。它确实适用于python 2.6,但在python 3.1上,我得到一个错误:

Traceback (most recent call last):
  File "unicode-file.py", line 15, in <module>
    old_file_write = file.write
NameError: name 'file' is not defined

unicode友好文件。open()替换

#!/usr/bin/python
import codecs, sys, types

# we save the file function handler because we want to override it
open_old = open

# on Python 3.x we overwrite write method in order to make it accept bytes in addition to str
old_file_write = file.write

class file():
    def write(self, d):
        if isinstance(d, types.bytes):
            self.buffer.write(d)
        else:
            old_file_write(d)

def open(filename, mode=None, bufsize=None):
    #try:
        # we read the first 4 bytes just to be sure we use the right encoding
        if(mode == "r"): # we are interested of detecting the mode only for read text
            f = open_old(filename, "rb")
            aBuf = f.read(4)
            if aBuf[:3] ==   '\xEF\xBB\xBF' :
                f = codecs.open(filename, mode, "utf_8")
                f.seek(3,0)
            elif aBuf[:4] == '\xFF\xFE\x00\x00':
                f = codecs.open(filename, mode, "utf_32_le")
                f.seek(4,0)
            elif aBuf[:4] == '\x00\x00\xFE\xFF': 
                f = codecs.open(filename, mode, "utf_32_be")
                f.seek(4,0)
            elif aBuf[:2] == '\xFF\xFE':
                f = codecs.open(filename, mode, "utf_16_le")
                f.seek(2,0)
            elif aBuf[:2] == '\xFE\xFF':
                f = codecs.open(filename, mode, "utf_16_be")
                f.seek(2,0)
            else:  # we assume that if there is no BOM, the encoding is UTF-8
                f.close()
                f = codecs.open(filename, mode, "utf-8")
                f.seek(0)
            return f
        else:
            return open_old(filename, mode, bufsize)

# now use the open(file, "r")

我已经润色了亚历克斯和索林的例子来研究python3和python2:

import codecs
import io

_boms = [
    (codecs.BOM_UTF8, 'utf-8-sig', 0),
    (codecs.BOM_UTF32_LE, 'utf-32le', 4),
    (codecs.BOM_UTF32_BE, 'utf-32be', 4),
    (codecs.BOM_UTF16_LE, 'utf-16le', 2),
    (codecs.BOM_UTF16_BE, 'utf-16be', 2)]


def read_unicode(file_path):
    with io.open(file_path, 'rb') as f:
        data = f.read(4)
    for bom, encoding, seek_to in _boms:
        if data.startswith(bom):
            break
    else:
        encoding, seek_to = 'utf-8', 0
    with io.open(file_path, 'r', encoding=encoding) as f:
        f.seek(seek_to)
        return f.read()