43 

    44 def split(stream):

    45     '''return an iterator of individual patches from a stream'''

    46     def isheader(line, inheader):

    47         if inheader and line[0] in (' ', '\t'):

    48             # continuation

    49             return True

    50         l = line.split(': ', 1)

    51         return len(l) == 2 and ' ' not in l[0]

    52 

    53     def chunk(lines):

    54         return cStringIO.StringIO(''.join(lines))

    55 

    56     def hgsplit(stream, cur):

    57         inheader = True

    58 

    59         for line in stream:

    60             if not line.strip():

    61                 inheader = False

    62             if not inheader and line.startswith('# HG changeset patch'):

    63                 yield chunk(cur)

    64                 cur = []

    65                 inheader = True

    66 

    67             cur.append(line)

    68 

    69         if cur:

    70             yield chunk(cur)

    71 

    72     def mboxsplit(stream, cur):

    73         for line in stream:

    74             if line.startswith('From '):

    75                 for c in split(chunk(cur[1:])):

    76                     yield c

    77                 cur = []

    78 

    79             cur.append(line)

    80 

    81         if cur:

    82             for c in split(chunk(cur[1:])):

    83                 yield c

    84 

    85     def mimesplit(stream, cur):

    86         def msgfp(m):

    87             fp = cStringIO.StringIO()

    88             g = email.Generator.Generator(fp, mangle_from_=False)

    89             g.flatten(m)

    90             fp.seek(0)

    91             return fp

    92 

    93         for line in stream:

    94             cur.append(line)

    95         c = chunk(cur)

    96 

    97         m = email.Parser.Parser().parse(c)

    98         if not m.is_multipart():

    99             yield msgfp(m)

   100         else:

   101             ok_types = ('text/plain', 'text/x-diff', 'text/x-patch')

   102             for part in m.walk():

   103                 ct = part.get_content_type()

   104                 if ct not in ok_types:

   105                     continue

   106                 yield msgfp(part)

   107 

   108     def headersplit(stream, cur):

   109         inheader = False

   110 

   111         for line in stream:

   112             if not inheader and isheader(line, inheader):

   113                 yield chunk(cur)

   114                 cur = []

   115                 inheader = True

   116             if inheader and not isheader(line, inheader):

   117                 inheader = False

   118 

   119             cur.append(line)

   120 

   121         if cur:

   122             yield chunk(cur)

   123 

   124     def remainder(cur):

   125         yield chunk(cur)

   126 

   127     class fiter(object):

   128         def __init__(self, fp):

   129             self.fp = fp

   130 

   131         def __iter__(self):

   132             return self

   133 

   134         def next(self):

   135             l = self.fp.readline()

   136             if not l:

   137                 raise StopIteration

   138             return l

   139 

   140     inheader = False

   141     cur = []

   142 

   143     mimeheaders = ['content-type']

   144 

   145     if not hasattr(stream, 'next'):

   146         # http responses, for example, have readline but not next

   147         stream = fiter(stream)

   148 

   149     for line in stream:

   150         cur.append(line)

   151         if line.startswith('# HG changeset patch'):

   152             return hgsplit(stream, cur)

   153         elif line.startswith('From '):

   154             return mboxsplit(stream, cur)

   155         elif isheader(line, inheader):

   156             inheader = True

   157             if line.split(':', 1)[0].lower() in mimeheaders:

   158                 # let email parser handle this

   159                 return mimesplit(stream, cur)

   160         elif inheader:

   161             # No evil headers seen, split by hand

   162             return headersplit(stream, cur)

   163         # Not enough info, keep reading

   164 

   165     # if we are here, we have a very plain patch

   166     return remainder(cur)