从电子邮件源码中提取文本，依赖于beautifulsoup

email2text.py

#!/usr/bin/python3

from codecs import getdecoder
from email import message_from_bytes
from bs4 import BeautifulSoup

ATTACHMENT = '<attachment with type %s>'
GARBAGE = '<cannot decode>'


def safe(func):
    def wrapper(*args, **kwargs):
        try:
            return func(*args, **kwargs)
        except:
            return GARBAGE
    return wrapper


def payload(m):
    encoding = m.get_content_charset() or 'utf8'
    try:
        getdecoder(encoding)
    except LookupError:
        encoding = 'utf8'
    try:
        s = m.get_payload(decode=True)
    except:
        return GARBAGE
    return s.decode(encoding, errors='ignore')


@safe
def text_plain(m):
    return payload(m)

@safe
def text_html(m):
    soup = BeautifulSoup(payload(m), 'lxml')
    return soup.text

@safe
def text_enriched(m):
    soup = BeautifulSoup(payload(m), 'lxml')
    return soup.text

@safe
def multipart_alternative(m):
    return msg2text(m.get_payload()[-1])

@safe
def multipart_mixed(m):
    return '\n'.join(map(msg2text, m.get_payload()))

@safe
def multipart_related(m):
    return '\n'.join(map(msg2text, m.get_payload()))

@safe
def multipart_signed(m):
    return msg2text(m.get_payload()[0])

@safe
def multipart_report(m):
    return msg2text(m.get_payload()[0])


def msg2text(m):
    t = m.get_content_type()
    if t == 'text/plain':
        return text_plain(m)
    elif t == 'text/html':
        return text_html(m)
    elif t == 'text/enriched':
        return text_enriched(m)
    elif t == 'multipart/alternative':
        return multipart_alternative(m)
    elif t == 'multipart/mixed':
        return multipart_mixed(m)
    elif t == 'multipart/related':
        return multipart_related(m)
    elif t == 'multipart/signed':
        return multipart_signed(m)
    elif t == 'multipart/report':
        return multipart_report(m)
    elif t.startswith('text/'):
        raise ValueError('Unknown MIME type: %s' % t)
    elif t.startswith('multipart/'):
        raise ValueError('Unknown MIME type: %s' % t)
    else:
        return ATTACHMENT % t

def email2text(f):
    """f should be opened with 'rb' mode!"""
    return msg2text(message_from_bytes(f.read()))


if __name__ == '__main__':
    from sys import argv, stderr
    if len(argv) == 1:
        print('Usage: %s <file>' % argv[0], file=stderr)
        print('You can also import this and use `email2text` function.',
                file=stderr)
        exit(1)
    for path in argv[1:]:
        if len(argv) > 2:
            print(path)
        with open(path, 'rb') as f:
            print(email2text(f))
        if len(argv) > 2:
            print()

么么馒头