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
24
from __future__ import absolute_import
27
from cStringIO import StringIO
36
# A RangeFile expects the following grammar (simplified to outline the
37
# assumptions we rely upon).
43
# whole_file: [content_length_header] data
45
# single_range: content_range_header data
47
# multiple_range: boundary_header boundary (content_range_header data boundary)+
49
class RangeFile(object):
50
"""File-like object that allow access to partial available data.
52
All accesses should happen sequentially since the acquisition occurs during
53
an http response reception (as sockets can't be seeked, we simulate the
54
seek by just reading and discarding the data).
56
The access pattern is defined by a set of ranges discovered as reading
57
progress. Only one range is available at a given time, so all accesses
58
should happen with monotonically increasing offsets.
61
# in _checked_read() below, we may have to discard several MB in the worst
62
# 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
64
# 8k chunks should be fine.
65
_discarded_buf_size = 8192
67
# maximum size of read requests -- used to avoid MemoryError issues in recv
68
_max_read_size = 512 * 1024
70
def __init__(self, path, infile):
73
:param path: File url, for error reports.
74
:param infile: File-like socket set at body start.
79
# When using multi parts response, this will be set with the headers
80
# associated with the range currently read.
82
# Default to the whole file of unspecified size
88
Dummy implementation for consistency with the 'file' API.
91
def set_range(self, start, size):
92
"""Change the range mapping"""
95
# Set the new _pos since that's what we want to expose
96
self._pos = self._start
98
def set_boundary(self, boundary):
99
"""Define the boundary used in a multi parts message.
101
The file should be at the beginning of the body, the first range
102
definition is read and taken into account.
104
self._boundary = boundary
105
# Decode the headers and setup the first range
107
self.read_range_definition()
109
def read_boundary(self):
110
"""Read the boundary headers defining a new range"""
111
boundary_line = '\r\n'
112
while boundary_line == '\r\n':
113
# RFC2616 19.2 Additional CRLFs may precede the first boundary
115
# To be on the safe side we allow it before any boundary line
116
boundary_line = self._file.readline()
118
if boundary_line == '':
119
# A timeout in the proxy server caused the response to end early.
120
# See launchpad bug 198646.
121
raise errors.HttpBoundaryMissing(
125
if boundary_line != '--' + self._boundary + '\r\n':
126
# rfc822.unquote() incorrectly unquotes strings enclosed in <>
127
# IIS 6 and 7 incorrectly wrap boundary strings in <>
128
# together they make a beautiful bug, which we will be gracious
130
if (self._unquote_boundary(boundary_line) !=
131
'--' + self._boundary + '\r\n'):
132
raise errors.InvalidHttpResponse(
134
"Expected a boundary (%s) line, got '%s'"
135
% (self._boundary, boundary_line))
137
def _unquote_boundary(self, b):
138
return b[:2] + rfc822.unquote(b[2:-2]) + b[-2:]
140
def read_range_definition(self):
141
"""Read a new range definition in a multi parts message.
143
Parse the headers including the empty line following them so that we
144
are ready to read the data itself.
146
self._headers = httplib.HTTPMessage(self._file, seekable=0)
147
# Extract the range definition
148
content_range = self._headers.getheader('content-range', None)
149
if content_range is None:
150
raise errors.InvalidHttpResponse(
152
'Content-Range header missing in a multi-part response')
153
self.set_range_from_header(content_range)
155
def set_range_from_header(self, content_range):
156
"""Helper to set the new range from its description in the headers"""
158
rtype, values = content_range.split()
160
raise errors.InvalidHttpRange(self._path, content_range,
163
raise errors.InvalidHttpRange(self._path, content_range,
164
"Unsupported range type '%s'" % rtype)
166
# We don't need total, but note that it may be either the file size
167
# or '*' if the server can't or doesn't want to return the file
169
start_end, total = values.split('/')
170
start, end = start_end.split('-')
174
raise errors.InvalidHttpRange(self._path, content_range,
175
'Invalid range values')
176
size = end - start + 1
178
raise errors.InvalidHttpRange(self._path, content_range,
179
'Invalid range, size <= 0')
180
self.set_range(start, size)
182
def _checked_read(self, size):
183
"""Read the file checking for short reads.
185
The data read is discarded along the way.
190
data = self._file.read(min(remaining, self._discarded_buf_size))
191
remaining -= len(data)
193
raise errors.ShortReadvError(self._path, pos, size,
197
def _seek_to_next_range(self):
198
# We will cross range boundaries
199
if self._boundary is None:
200
# If we don't have a boundary, we can't find another range
201
raise errors.InvalidRange(self._path, self._pos,
202
"Range (%s, %s) exhausted"
203
% (self._start, self._size))
205
self.read_range_definition()
207
def read(self, size=-1):
208
"""Read size bytes from the current position in the file.
210
Reading across ranges is not supported. We rely on the underlying http
211
client to clean the socket if we leave bytes unread. This may occur for
212
the final boundary line of a multipart response or for any range
213
request not entirely consumed by the client (due to offset coalescing)
215
:param size: The number of bytes to read. Leave unspecified or pass
219
and self._pos == self._start + self._size):
223
self._seek_to_next_range()
224
elif self._pos < self._start:
225
raise errors.InvalidRange(
226
self._path, self._pos,
227
"Can't read %s bytes before range (%s, %s)"
228
% (size, self._start, self._size))
230
if size > 0 and self._pos + size > self._start + self._size:
231
raise errors.InvalidRange(
232
self._path, self._pos,
233
"Can't read %s bytes across range (%s, %s)"
234
% (size, self._start, self._size))
236
# read data from file
240
# Don't read past the range definition
241
limited = self._start + self._size - self._pos
243
limited = min(limited, size)
244
osutils.pumpfile(self._file, buffer, limited, self._max_read_size)
245
data = buffer.getvalue()
247
# Update _pos respecting the data effectively read
248
self._pos += len(data)
251
def seek(self, offset, whence=0):
252
start_pos = self._pos
256
final_pos = start_pos + offset
259
final_pos = self._start + self._size + offset # offset < 0
261
raise errors.InvalidRange(
262
self._path, self._pos,
263
"RangeFile: can't seek from end while size is unknown")
265
raise ValueError("Invalid value %s for whence." % whence)
267
if final_pos < self._pos:
268
# Can't seek backwards
269
raise errors.InvalidRange(
270
self._path, self._pos,
271
'RangeFile: trying to seek backwards to %s' % final_pos)
274
cur_limit = self._start + self._size
275
while final_pos > cur_limit:
276
# We will cross range boundaries
277
remain = cur_limit - self._pos
279
# Finish reading the current range
280
self._checked_read(remain)
281
self._seek_to_next_range()
282
cur_limit = self._start + self._size
284
size = final_pos - self._pos
285
if size > 0: # size can be < 0 if we crossed a range boundary
286
# We don't need the data, just read it and throw it away
287
self._checked_read(size)
293
def handle_response(url, code, msg, data):
294
"""Interpret the code & headers and wrap the provided data in a RangeFile.
296
This is a factory method which returns an appropriate RangeFile based on
297
the code & headers it's given.
299
:param url: The url being processed. Mostly for error reporting
300
:param code: The integer HTTP response code
301
:param msg: An HTTPMessage containing the headers for the response
302
:param data: A file-like object that can be read() to get the
304
:return: A file-like object that can seek()+read() the
305
ranges indicated by the headers.
307
rfile = RangeFile(url, data)
310
size = msg.getheader('content-length', None)
315
rfile.set_range(0, size)
317
content_type = msg.getheader('content-type', None)
318
if content_type is None:
319
# When there is no content-type header we treat the response as
320
# being of type 'application/octet-stream' as per RFC2616 section
322
# Therefore it is obviously not multipart
323
content_type = 'application/octet-stream'
326
is_multipart = (msg.getmaintype() == 'multipart'
327
and msg.getsubtype() == 'byteranges')
330
# Full fledged multipart response
331
rfile.set_boundary(msg.getparam('boundary'))
333
# A response to a range request, but not multipart
334
content_range = msg.getheader('content-range', None)
335
if content_range is None:
336
raise errors.InvalidHttpResponse(url,
337
'Missing the Content-Range header in a 206 range response')
338
rfile.set_range_from_header(content_range)
340
raise errors.InvalidHttpResponse(url,
341
'Unknown response code %s' % code)
added
removed
bzrlib/transport/http/response.py
