26
from cStringIO import StringIO
25
from io import BytesIO
27
import http.client as http_client
28
import email.utils as email_utils
36
class ResponseFile(object):
37
"""A wrapper around the http socket containing the result of a GET request.
39
Only read() and seek() (forward) are supported.
43
def __init__(self, path, infile):
46
:param path: File url, for error reports.
48
:param infile: File-like socket set at body start.
57
Dummy implementation for consistency with the 'file' API.
63
def __exit__(self, exc_type, exc_val, exc_tb):
64
return False # propogate exceptions.
66
def read(self, size=None):
67
"""Read size bytes from the current position in the file.
69
:param size: The number of bytes to read. Leave unspecified or pass
72
data = self._file.read(size)
73
self._pos += len(data)
77
data = self._file.readline()
78
self._pos += len(data)
83
line = self.readline()
91
def seek(self, offset, whence=os.SEEK_SET):
92
if whence == os.SEEK_SET:
93
if offset < self._pos:
95
"Can't seek backwards, pos: %s, offset: %s"
96
% (self._pos, offset))
97
to_discard = offset - self._pos
98
elif whence == os.SEEK_CUR:
101
raise AssertionError("Can't seek backwards")
103
# Just discard the unwanted bytes
104
self.read(to_discard)
36
106
# A RangeFile expects the following grammar (simplified to outline the
37
107
# assumptions we rely upon).
41
110
# | multiple_range
43
# whole_file: [content_length_header] data
45
112
# single_range: content_range_header data
47
114
# multiple_range: boundary_header boundary (content_range_header data boundary)+
49
class RangeFile(object):
117
class RangeFile(ResponseFile):
50
118
"""File-like object that allow access to partial available data.
52
120
All accesses should happen sequentially since the acquisition occurs during
61
129
# in _checked_read() below, we may have to discard several MB in the worst
62
130
# case. To avoid buffering that much, we read and discard by chunks
63
# instead. The underlying file is either a socket or a StringIO, so reading
131
# instead. The underlying file is either a socket or a BytesIO, so reading
64
132
# 8k chunks should be fine.
65
133
_discarded_buf_size = 8192
73
141
:param path: File url, for error reports.
74
143
:param infile: File-like socket set at body start.
145
super(RangeFile, self).__init__(path, infile)
78
146
self._boundary = None
79
147
# When using multi parts response, this will be set with the headers
80
148
# associated with the range currently read.
95
163
The file should be at the beginning of the body, the first range
96
164
definition is read and taken into account.
166
if not isinstance(boundary, bytes):
167
raise TypeError(boundary)
98
168
self._boundary = boundary
99
169
# Decode the headers and setup the first range
100
170
self.read_boundary()
103
173
def read_boundary(self):
104
174
"""Read the boundary headers defining a new range"""
105
boundary_line = '\r\n'
106
while boundary_line == '\r\n':
175
boundary_line = b'\r\n'
176
while boundary_line == b'\r\n':
107
177
# RFC2616 19.2 Additional CRLFs may precede the first boundary
109
179
# To be on the safe side we allow it before any boundary line
110
180
boundary_line = self._file.readline()
112
if boundary_line != '--' + self._boundary + '\r\n':
113
# rfc822.unquote() incorrectly unquotes strings enclosed in <>
182
if boundary_line == b'':
183
# A timeout in the proxy server caused the response to end early.
184
# See launchpad bug 198646.
185
raise errors.HttpBoundaryMissing(
189
if boundary_line != b'--' + self._boundary + b'\r\n':
190
# email_utils.unquote() incorrectly unquotes strings enclosed in <>
114
191
# IIS 6 and 7 incorrectly wrap boundary strings in <>
115
192
# together they make a beautiful bug, which we will be gracious
117
194
if (self._unquote_boundary(boundary_line) !=
118
'--' + self._boundary + '\r\n'):
195
b'--' + self._boundary + b'\r\n'):
119
196
raise errors.InvalidHttpResponse(
121
198
"Expected a boundary (%s) line, got '%s'"
122
199
% (self._boundary, boundary_line))
124
201
def _unquote_boundary(self, b):
125
return b[:2] + rfc822.unquote(b[2:-2]) + b[-2:]
202
return b[:2] + email_utils.unquote(b[2:-2].decode('ascii')).encode('ascii') + b[-2:]
127
204
def read_range_definition(self):
128
205
"""Read a new range definition in a multi parts message.
130
207
Parse the headers including the empty line following them so that we
131
208
are ready to read the data itself.
133
self._headers = httplib.HTTPMessage(self._file, seekable=0)
210
self._headers = http_client.parse_headers(self._file)
134
211
# Extract the range definition
135
content_range = self._headers.getheader('content-range', None)
212
content_range = self._headers.get('content-range', None)
136
213
if content_range is None:
137
214
raise errors.InvalidHttpResponse(
221
298
% (size, self._start, self._size))
223
300
# read data from file
226
303
if self._size > 0:
227
304
# Don't read past the range definition
228
305
limited = self._start + self._size - self._pos
230
307
limited = min(limited, size)
231
osutils.pumpfile(self._file, buffer, limited, self._max_read_size)
232
data = buffer.getvalue()
308
osutils.pumpfile(self._file, buf, limited, self._max_read_size)
309
data = buf.getvalue()
234
311
# Update _pos respecting the data effectively read
235
312
self._pos += len(data)
269
346
cur_limit = self._start + self._size
271
348
size = final_pos - self._pos
272
if size > 0: # size can be < 0 if we crossed a range boundary
349
if size > 0: # size can be < 0 if we crossed a range boundary
273
350
# We don't need the data, just read it and throw it away
274
351
self._checked_read(size)
280
def handle_response(url, code, msg, data):
357
def handle_response(url, code, getheader, data):
281
358
"""Interpret the code & headers and wrap the provided data in a RangeFile.
283
360
This is a factory method which returns an appropriate RangeFile based on
286
363
:param url: The url being processed. Mostly for error reporting
287
364
:param code: The integer HTTP response code
288
:param msg: An HTTPMessage containing the headers for the response
365
:param getheader: Function for retrieving header
289
366
:param data: A file-like object that can be read() to get the
291
368
:return: A file-like object that can seek()+read() the
292
369
ranges indicated by the headers.
294
rfile = RangeFile(url, data)
297
size = msg.getheader('content-length', None)
302
rfile.set_range(0, size)
373
rfile = ResponseFile(url, data)
303
374
elif code == 206:
304
content_type = msg.getheader('content-type', None)
305
if content_type is None:
306
# When there is no content-type header we treat the response as
307
# being of type 'application/octet-stream' as per RFC2616 section
309
# Therefore it is obviously not multipart
310
content_type = 'application/octet-stream'
313
is_multipart = (msg.getmaintype() == 'multipart'
314
and msg.getsubtype() == 'byteranges')
317
# Full fledged multipart response
318
rfile.set_boundary(msg.getparam('boundary'))
375
rfile = RangeFile(url, data)
376
# When there is no content-type header we treat the response as
377
# being of type 'application/octet-stream' as per RFC2616 section
379
# Therefore it is obviously not multipart
380
content_type = getheader('content-type', 'application/octet-stream')
381
mimetype, options = cgi.parse_header(content_type)
382
if mimetype == 'multipart/byteranges':
383
rfile.set_boundary(options['boundary'].encode('ascii'))
320
385
# A response to a range request, but not multipart
321
content_range = msg.getheader('content-range', None)
386
content_range = getheader('content-range', None)
322
387
if content_range is None:
323
raise errors.InvalidHttpResponse(url,
324
'Missing the Content-Range header in a 206 range response')
388
raise errors.InvalidHttpResponse(
389
url, 'Missing the Content-Range header in a 206 range response')
325
390
rfile.set_range_from_header(content_range)
327
392
raise errors.InvalidHttpResponse(url,
328
393
'Unknown response code %s' % code)