26
from cStringIO import StringIO
24
from __future__ import absolute_import
29
import http.client as http_client
30
except ImportError: # python < 3
31
import httplib as http_client
33
import email.utils as email_utils
34
except ImportError: # python < 3
35
import rfc822 as email_utils
41
from ...sixish import (
47
class ResponseFile(object):
48
"""A wrapper around the http socket containing the result of a GET request.
50
Only read() and seek() (forward) are supported.
54
def __init__(self, path, infile):
57
:param path: File url, for error reports.
59
:param infile: File-like socket set at body start.
68
Dummy implementation for consistency with the 'file' API.
74
def __exit__(self, exc_type, exc_val, exc_tb):
75
return False # propogate exceptions.
77
def read(self, size=None):
78
"""Read size bytes from the current position in the file.
80
:param size: The number of bytes to read. Leave unspecified or pass
83
if size is None and not PY3:
85
data = self._file.read(size)
86
self._pos += len(data)
90
data = self._file.readline()
91
self._pos += len(data)
94
def readlines(self, size=None):
95
data = self._file.readlines()
96
self._pos += sum(map(len, data))
101
line = self.readline()
109
def seek(self, offset, whence=os.SEEK_SET):
110
if whence == os.SEEK_SET:
111
if offset < self._pos:
112
raise AssertionError(
113
"Can't seek backwards, pos: %s, offset: %s"
114
% (self._pos, offset))
115
to_discard = offset - self._pos
116
elif whence == os.SEEK_CUR:
119
raise AssertionError("Can't seek backwards")
121
# Just discard the unwanted bytes
122
self.read(to_discard)
36
124
# A RangeFile expects the following grammar (simplified to outline the
37
125
# assumptions we rely upon).
41
128
# | multiple_range
43
# whole_file: [content_length_header] data
45
130
# single_range: content_range_header data
47
132
# multiple_range: boundary_header boundary (content_range_header data boundary)+
49
class RangeFile(object):
135
class RangeFile(ResponseFile):
50
136
"""File-like object that allow access to partial available data.
52
138
All accesses should happen sequentially since the acquisition occurs during
61
147
# in _checked_read() below, we may have to discard several MB in the worst
62
148
# 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
149
# instead. The underlying file is either a socket or a BytesIO, so reading
64
150
# 8k chunks should be fine.
65
151
_discarded_buf_size = 8192
73
159
:param path: File url, for error reports.
74
161
:param infile: File-like socket set at body start.
163
super(RangeFile, self).__init__(path, infile)
78
164
self._boundary = None
79
165
# When using multi parts response, this will be set with the headers
80
166
# associated with the range currently read.
95
181
The file should be at the beginning of the body, the first range
96
182
definition is read and taken into account.
184
if not isinstance(boundary, bytes):
185
raise TypeError(boundary)
98
186
self._boundary = boundary
99
187
# Decode the headers and setup the first range
100
188
self.read_boundary()
103
191
def read_boundary(self):
104
192
"""Read the boundary headers defining a new range"""
105
boundary_line = '\r\n'
106
while boundary_line == '\r\n':
193
boundary_line = b'\r\n'
194
while boundary_line == b'\r\n':
107
195
# RFC2616 19.2 Additional CRLFs may precede the first boundary
109
197
# To be on the safe side we allow it before any boundary line
110
198
boundary_line = self._file.readline()
112
if boundary_line != '--' + self._boundary + '\r\n':
113
# rfc822.unquote() incorrectly unquotes strings enclosed in <>
200
if boundary_line == b'':
201
# A timeout in the proxy server caused the response to end early.
202
# See launchpad bug 198646.
203
raise errors.HttpBoundaryMissing(
207
if boundary_line != b'--' + self._boundary + b'\r\n':
208
# email_utils.unquote() incorrectly unquotes strings enclosed in <>
114
209
# IIS 6 and 7 incorrectly wrap boundary strings in <>
115
210
# together they make a beautiful bug, which we will be gracious
117
212
if (self._unquote_boundary(boundary_line) !=
118
'--' + self._boundary + '\r\n'):
213
b'--' + self._boundary + b'\r\n'):
119
214
raise errors.InvalidHttpResponse(
121
216
"Expected a boundary (%s) line, got '%s'"
122
217
% (self._boundary, boundary_line))
124
219
def _unquote_boundary(self, b):
125
return b[:2] + rfc822.unquote(b[2:-2]) + b[-2:]
220
return b[:2] + email_utils.unquote(b[2:-2].decode('ascii')).encode('ascii') + b[-2:]
127
222
def read_range_definition(self):
128
223
"""Read a new range definition in a multi parts message.
130
225
Parse the headers including the empty line following them so that we
131
226
are ready to read the data itself.
133
self._headers = httplib.HTTPMessage(self._file, seekable=0)
229
self._headers = http_client.parse_headers(self._file)
231
self._headers = http_client.HTTPMessage(self._file, seekable=0)
134
232
# Extract the range definition
135
content_range = self._headers.getheader('content-range', None)
233
content_range = self._headers.get('content-range', None)
136
234
if content_range is None:
137
235
raise errors.InvalidHttpResponse(
221
319
% (size, self._start, self._size))
223
321
# read data from file
226
324
if self._size > 0:
227
325
# Don't read past the range definition
228
326
limited = self._start + self._size - self._pos
230
328
limited = min(limited, size)
231
osutils.pumpfile(self._file, buffer, limited, self._max_read_size)
232
data = buffer.getvalue()
329
osutils.pumpfile(self._file, buf, limited, self._max_read_size)
330
data = buf.getvalue()
234
332
# Update _pos respecting the data effectively read
235
333
self._pos += len(data)
280
def handle_response(url, code, msg, data):
378
def handle_response(url, code, getheader, data):
281
379
"""Interpret the code & headers and wrap the provided data in a RangeFile.
283
381
This is a factory method which returns an appropriate RangeFile based on
286
384
:param url: The url being processed. Mostly for error reporting
287
385
:param code: The integer HTTP response code
288
:param msg: An HTTPMessage containing the headers for the response
386
:param getheader: Function for retrieving header
289
387
:param data: A file-like object that can be read() to get the
291
389
:return: A file-like object that can seek()+read() the
292
390
ranges indicated by the headers.
294
rfile = RangeFile(url, data)
297
size = msg.getheader('content-length', None)
302
rfile.set_range(0, size)
394
rfile = ResponseFile(url, data)
303
395
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'))
396
rfile = RangeFile(url, data)
397
# When there is no content-type header we treat the response as
398
# being of type 'application/octet-stream' as per RFC2616 section
400
# Therefore it is obviously not multipart
401
content_type = getheader('content-type', 'application/octet-stream')
402
mimetype, options = cgi.parse_header(content_type)
403
if mimetype == 'multipart/byteranges':
404
rfile.set_boundary(options['boundary'].encode('ascii'))
320
406
# A response to a range request, but not multipart
321
content_range = msg.getheader('content-range', None)
407
content_range = getheader('content-range', None)
322
408
if content_range is None:
323
raise errors.InvalidHttpResponse(url,
324
'Missing the Content-Range header in a 206 range response')
409
raise errors.InvalidHttpResponse(
410
url, 'Missing the Content-Range header in a 206 range response')
325
411
rfile.set_range_from_header(content_range)
327
raise errors.InvalidHttpResponse(url,
328
'Unknown response code %s' % code)
413
raise errors.UnexpectedHttpStatus(url, code)