1
# Copyright (C) 2006-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""Handlers for HTTP Responses.
19
The purpose of these classes is to provide a uniform interface for clients
20
to standard HTTP responses, single range responses and multipart range
26
from cStringIO import StringIO
35
# A RangeFile expects the following grammar (simplified to outline the
36
# assumptions we rely upon).
42
# whole_file: [content_length_header] data
44
# single_range: content_range_header data
46
# multiple_range: boundary_header boundary (content_range_header data boundary)+
48
class RangeFile(object):
49
"""File-like object that allow access to partial available data.
51
All accesses should happen sequentially since the acquisition occurs during
52
an http response reception (as sockets can't be seeked, we simulate the
53
seek by just reading and discarding the data).
55
The access pattern is defined by a set of ranges discovered as reading
56
progress. Only one range is available at a given time, so all accesses
57
should happen with monotonically increasing offsets.
60
# in _checked_read() below, we may have to discard several MB in the worst
61
# case. To avoid buffering that much, we read and discard by chunks
62
# instead. The underlying file is either a socket or a StringIO, so reading
63
# 8k chunks should be fine.
64
_discarded_buf_size = 8192
66
# maximum size of read requests -- used to avoid MemoryError issues in recv
67
_max_read_size = 512 * 1024
69
def __init__(self, path, infile):
72
:param path: File url, for error reports.
73
:param infile: File-like socket set at body start.
78
# When using multi parts response, this will be set with the headers
79
# associated with the range currently read.
81
# Default to the whole file of unspecified size
84
def set_range(self, start, size):
85
"""Change the range mapping"""
88
# Set the new _pos since that's what we want to expose
89
self._pos = self._start
91
def set_boundary(self, boundary):
92
"""Define the boundary used in a multi parts message.
94
The file should be at the beginning of the body, the first range
95
definition is read and taken into account.
97
self._boundary = boundary
98
# Decode the headers and setup the first range
100
self.read_range_definition()
102
def read_boundary(self):
103
"""Read the boundary headers defining a new range"""
104
boundary_line = '\r\n'
105
while boundary_line == '\r\n':
106
# RFC2616 19.2 Additional CRLFs may precede the first boundary
108
# To be on the safe side we allow it before any boundary line
109
boundary_line = self._file.readline()
111
if boundary_line == '':
112
# A timeout in the proxy server caused the response to end early.
113
# See launchpad bug 198646.
114
raise errors.HttpBoundaryMissing(
118
if boundary_line != '--' + self._boundary + '\r\n':
119
# rfc822.unquote() incorrectly unquotes strings enclosed in <>
120
# IIS 6 and 7 incorrectly wrap boundary strings in <>
121
# together they make a beautiful bug, which we will be gracious
123
if (self._unquote_boundary(boundary_line) !=
124
'--' + self._boundary + '\r\n'):
125
raise errors.InvalidHttpResponse(
127
"Expected a boundary (%s) line, got '%s'"
128
% (self._boundary, boundary_line))
130
def _unquote_boundary(self, b):
131
return b[:2] + rfc822.unquote(b[2:-2]) + b[-2:]
133
def read_range_definition(self):
134
"""Read a new range definition in a multi parts message.
136
Parse the headers including the empty line following them so that we
137
are ready to read the data itself.
139
self._headers = httplib.HTTPMessage(self._file, seekable=0)
140
# Extract the range definition
141
content_range = self._headers.getheader('content-range', None)
142
if content_range is None:
143
raise errors.InvalidHttpResponse(
145
'Content-Range header missing in a multi-part response')
146
self.set_range_from_header(content_range)
148
def set_range_from_header(self, content_range):
149
"""Helper to set the new range from its description in the headers"""
151
rtype, values = content_range.split()
153
raise errors.InvalidHttpRange(self._path, content_range,
156
raise errors.InvalidHttpRange(self._path, content_range,
157
"Unsupported range type '%s'" % rtype)
159
# We don't need total, but note that it may be either the file size
160
# or '*' if the server can't or doesn't want to return the file
162
start_end, total = values.split('/')
163
start, end = start_end.split('-')
167
raise errors.InvalidHttpRange(self._path, content_range,
168
'Invalid range values')
169
size = end - start + 1
171
raise errors.InvalidHttpRange(self._path, content_range,
172
'Invalid range, size <= 0')
173
self.set_range(start, size)
175
def _checked_read(self, size):
176
"""Read the file checking for short reads.
178
The data read is discarded along the way.
183
data = self._file.read(min(remaining, self._discarded_buf_size))
184
remaining -= len(data)
186
raise errors.ShortReadvError(self._path, pos, size,
190
def _seek_to_next_range(self):
191
# We will cross range boundaries
192
if self._boundary is None:
193
# If we don't have a boundary, we can't find another range
194
raise errors.InvalidRange(self._path, self._pos,
195
"Range (%s, %s) exhausted"
196
% (self._start, self._size))
198
self.read_range_definition()
200
def read(self, size=-1):
201
"""Read size bytes from the current position in the file.
203
Reading across ranges is not supported. We rely on the underlying http
204
client to clean the socket if we leave bytes unread. This may occur for
205
the final boundary line of a multipart response or for any range
206
request not entirely consumed by the client (due to offset coalescing)
208
:param size: The number of bytes to read. Leave unspecified or pass
212
and self._pos == self._start + self._size):
216
self._seek_to_next_range()
217
elif self._pos < self._start:
218
raise errors.InvalidRange(
219
self._path, self._pos,
220
"Can't read %s bytes before range (%s, %s)"
221
% (size, self._start, self._size))
223
if size > 0 and self._pos + size > self._start + self._size:
224
raise errors.InvalidRange(
225
self._path, self._pos,
226
"Can't read %s bytes across range (%s, %s)"
227
% (size, self._start, self._size))
229
# read data from file
233
# Don't read past the range definition
234
limited = self._start + self._size - self._pos
236
limited = min(limited, size)
237
osutils.pumpfile(self._file, buffer, limited, self._max_read_size)
238
data = buffer.getvalue()
240
# Update _pos respecting the data effectively read
241
self._pos += len(data)
244
def seek(self, offset, whence=0):
245
start_pos = self._pos
249
final_pos = start_pos + offset
252
final_pos = self._start + self._size + offset # offset < 0
254
raise errors.InvalidRange(
255
self._path, self._pos,
256
"RangeFile: can't seek from end while size is unknown")
258
raise ValueError("Invalid value %s for whence." % whence)
260
if final_pos < self._pos:
261
# Can't seek backwards
262
raise errors.InvalidRange(
263
self._path, self._pos,
264
'RangeFile: trying to seek backwards to %s' % final_pos)
267
cur_limit = self._start + self._size
268
while final_pos > cur_limit:
269
# We will cross range boundaries
270
remain = cur_limit - self._pos
272
# Finish reading the current range
273
self._checked_read(remain)
274
self._seek_to_next_range()
275
cur_limit = self._start + self._size
277
size = final_pos - self._pos
278
if size > 0: # size can be < 0 if we crossed a range boundary
279
# We don't need the data, just read it and throw it away
280
self._checked_read(size)
286
def handle_response(url, code, msg, data):
287
"""Interpret the code & headers and wrap the provided data in a RangeFile.
289
This is a factory method which returns an appropriate RangeFile based on
290
the code & headers it's given.
292
:param url: The url being processed. Mostly for error reporting
293
:param code: The integer HTTP response code
294
:param msg: An HTTPMessage containing the headers for the response
295
:param data: A file-like object that can be read() to get the
297
:return: A file-like object that can seek()+read() the
298
ranges indicated by the headers.
300
rfile = RangeFile(url, data)
303
size = msg.getheader('content-length', None)
308
rfile.set_range(0, size)
310
content_type = msg.getheader('content-type', None)
311
if content_type is None:
312
# When there is no content-type header we treat the response as
313
# being of type 'application/octet-stream' as per RFC2616 section
315
# Therefore it is obviously not multipart
316
content_type = 'application/octet-stream'
319
is_multipart = (msg.getmaintype() == 'multipart'
320
and msg.getsubtype() == 'byteranges')
323
# Full fledged multipart response
324
rfile.set_boundary(msg.getparam('boundary'))
326
# A response to a range request, but not multipart
327
content_range = msg.getheader('content-range', None)
328
if content_range is None:
329
raise errors.InvalidHttpResponse(url,
330
'Missing the Content-Range header in a 206 range response')
331
rfile.set_range_from_header(content_range)
333
raise errors.InvalidHttpResponse(url,
334
'Unknown response code %s' % code)
added
removed
bzrlib/transport/http/response.py
