# -*- coding: utf-8 -
#
# This file is part of restkit released under the MIT license.
# See the NOTICE for more information.
import base64
import errno
import logging
import os
import time
import socket
import ssl
import traceback
import types
import urlparse
try:
from http_parser.http import HttpStream, BadStatusLine
from http_parser.reader import SocketReader
except ImportError:
raise ImportError("""http-parser isn't installed or out of data.
pip install http-parser""")
from restkit import __version__
from restkit.conn import Connection
from restkit.errors import RequestError, RequestTimeout, RedirectLimit, \
NoMoreData, ProxyError
from restkit.session import get_session
from restkit.util import parse_netloc, rewrite_location, to_bytestring
from restkit.wrappers import Request, Response
MAX_CLIENT_TIMEOUT=300
MAX_CLIENT_CONNECTIONS = 5
MAX_CLIENT_TRIES =3
CLIENT_WAIT_TRIES = 0.3
MAX_FOLLOW_REDIRECTS = 5
USER_AGENT = "restkit/%s" % __version__
log = logging.getLogger(__name__)
class Client(object):
""" A client handle a connection at a time. A client is threadsafe,
but an handled shouldn't be shared between threads. All connections
are shared between threads via a pool.
>>> from restkit import *
>>> c = Client()
>>> r = c.request("http://google.com")
r>>> r.status
'301 Moved Permanently'
>>> r.body_string()
'\n301 Moved\n301 Moved
\nThe document has moved\nhere.\r\n\r\n'
>>> c.follow_redirect = True
>>> r = c.request("http://google.com")
>>> r.status
'200 OK'
"""
version = (1, 1)
response_class=Response
def __init__(self,
follow_redirect=False,
force_follow_redirect=False,
max_follow_redirect=MAX_FOLLOW_REDIRECTS,
filters=None,
decompress=True,
max_status_line_garbage=None,
max_header_count=0,
pool=None,
response_class=None,
timeout=None,
use_proxy=False,
max_tries=3,
wait_tries=0.3,
max_conn=150,
pool_size=10,
backend="thread",
**ssl_args):
"""
Client parameters
~~~~~~~~~~~~~~~~~
:param follow_redirect: follow redirection, by default False
:param max_ollow_redirect: number of redirections available
:filters: http filters to pass
:param decompress: allows the client to decompress the response
body
:param max_status_line_garbage: defines the maximum number of ignorable
lines before we expect a HTTP response's status line. With
HTTP/1.1 persistent connections, the problem arises that broken
scripts could return a wrong Content-Length (there are more
bytes sent than specified). Unfortunately, in some cases, this
cannot be detected after the bad response, but only before the
next one. So the client is abble to skip bad lines using this
limit. 0 disable garbage collection, None means unlimited number
of tries.
:param max_header_count: determines the maximum HTTP header count
allowed. by default no limit.
:param manager: the manager to use. By default we use the global
one.
:parama response_class: the response class to use
:param timeout: the default timeout of the connection
(SO_TIMEOUT)
:param max_tries: the number of tries before we give up a
connection
:param wait_tries: number of time we wait between each tries.
:param ssl_args: named argument, see ssl module for more
informations
"""
self.follow_redirect = follow_redirect
self.force_follow_redirect = force_follow_redirect
self.max_follow_redirect = max_follow_redirect
self.decompress = decompress
self.filters = filters or []
self.max_status_line_garbage = max_status_line_garbage
self.max_header_count = max_header_count
self.use_proxy = use_proxy
self.request_filters = []
self.response_filters = []
self.load_filters()
# set manager
session_options = dict(
retry_delay=wait_tries,
max_size = pool_size,
max_conn = max_conn,
retry_max = max_tries,
timeout = timeout)
if pool is None:
pool = get_session(backend, **session_options)
self._pool = pool
self.backend = backend
# change default response class
if response_class is not None:
self.response_class = response_class
self.max_tries = max_tries
self.wait_tries = wait_tries
self.max_conn = max_conn
self.pool_size = pool_size
self.timeout = timeout
self._nb_redirections = self.max_follow_redirect
self._url = None
self._initial_url = None
self._write_cb = None
self._headers = None
self._sock_key = None
self._sock = None
self._original = None
self.method = 'GET'
self.body = None
self.ssl_args = ssl_args or {}
def load_filters(self):
""" Populate filters from self.filters.
Must be called each time self.filters is updated.
"""
for f in self.filters:
if hasattr(f, "on_request"):
self.request_filters.append(f)
if hasattr(f, "on_response"):
self.response_filters.append(f)
def get_connection(self, request):
""" get a connection from the pool or create new one. """
addr = parse_netloc(request.parsed_url)
is_ssl = request.is_ssl()
extra_headers = []
conn = None
if self.use_proxy:
conn = self.proxy_connection(request,
addr, is_ssl)
if not conn:
conn = self._pool.get(host=addr[0], port=addr[1],
pool=self._pool, is_ssl=is_ssl,
extra_headers=extra_headers, **self.ssl_args)
return conn
def proxy_connection(self, request, req_addr, is_ssl):
""" do the proxy connection """
proxy_settings = os.environ.get('%s_proxy' %
request.parsed_url.scheme)
if proxy_settings and proxy_settings is not None:
request.is_proxied = True
proxy_settings, proxy_auth = _get_proxy_auth(proxy_settings)
addr = parse_netloc(urlparse.urlparse(proxy_settings))
if is_ssl:
if proxy_auth:
proxy_auth = 'Proxy-authorization: %s' % proxy_auth
proxy_connect = 'CONNECT %s:%s HTTP/1.0\r\n' % req_addr
user_agent = request.headers.iget('user_agent')
if not user_agent:
user_agent = "User-Agent: restkit/%s\r\n" % __version__
proxy_pieces = '%s%s%s\r\n' % (proxy_connect, proxy_auth,
user_agent)
conn = self._pool.get(host=addr[0], port=addr[1],
pool=self._pool, is_ssl=is_ssl,
extra_headers=[], **self.ssl_args)
conn.send(proxy_pieces)
p = HttpStream(SocketReader(conn.socket()), kind=1,
decompress=True)
if p.status_code != 200:
raise ProxyError("Tunnel connection failed: %d %s" %
(p.status_code(), p.body_string()))
# read body
p.body_string()
else:
headers = []
if proxy_auth:
headers = [('Proxy-authorization', proxy_auth)]
conn = self._pool.get(host=addr[0], port=addr[1],
pool=self._pool, is_ssl=False,
extra_headers=[], **self.ssl_args)
return conn
return
def make_headers_string(self, request, extra_headers=None):
""" create final header string """
headers = request.headers.copy()
if extra_headers is not None:
for k, v in extra_headers:
headers[k] = v
if not request.body and request.method in ('POST', 'PUT',):
headers['Content-Length'] = 0
if self.version == (1,1):
httpver = "HTTP/1.1"
else:
httpver = "HTTP/1.0"
ua = headers.iget('user_agent')
if not ua:
ua = USER_AGENT
host = request.host
accept_encoding = headers.iget('accept-encoding')
if not accept_encoding:
accept_encoding = 'identity'
if request.is_proxied:
full_path = ("https://" if request.is_ssl() else "http://") + request.host + request.path
else:
full_path = request.path
lheaders = [
"%s %s %s\r\n" % (request.method, full_path, httpver),
"Host: %s\r\n" % host,
"User-Agent: %s\r\n" % ua,
"Accept-Encoding: %s\r\n" % accept_encoding
]
lheaders.extend(["%s: %s\r\n" % (k, str(v)) for k, v in \
headers.items() if k.lower() not in \
('user-agent', 'host', 'accept-encoding',)])
if log.isEnabledFor(logging.DEBUG):
log.debug("Send headers: %s" % lheaders)
return "%s\r\n" % "".join(lheaders)
def perform(self, request):
""" perform the request. If an error happen it will first try to
restart it """
if log.isEnabledFor(logging.DEBUG):
log.debug("Start to perform request: %s %s %s" %
(request.host, request.method, request.path))
tries = 0
while True:
conn = None
try:
# get or create a connection to the remote host
conn = self.get_connection(request)
# send headers
msg = self.make_headers_string(request,
conn.extra_headers)
# send body
if request.body is not None:
chunked = request.is_chunked()
if request.headers.iget('content-length') is None and \
not chunked:
raise RequestError(
"Can't determine content length and " +
"Transfer-Encoding header is not chunked")
# handle 100-Continue status
# http://www.w3.org/Protocols/rfc2616/rfc2616-sec8.html#sec8.2.3
hdr_expect = request.headers.iget("expect")
if hdr_expect is not None and \
hdr_expect.lower() == "100-continue":
conn.send(msg)
msg = None
p = HttpStream(SocketReader(conn.socket()), kind=1,
decompress=True)
if p.status_code != 100:
self.reset_request()
if log.isEnabledFor(logging.DEBUG):
log.debug("return response class")
return self.response_class(conn, request, p)
chunked = request.is_chunked()
if log.isEnabledFor(logging.DEBUG):
log.debug("send body (chunked: %s)" % chunked)
if isinstance(request.body, types.StringTypes):
if msg is not None:
conn.send(msg + to_bytestring(request.body),
chunked)
else:
conn.send(to_bytestring(request.body), chunked)
else:
if msg is not None:
conn.send(msg)
if hasattr(request.body, 'read'):
if hasattr(request.body, 'seek'):
request.body.seek(0)
conn.sendfile(request.body, chunked)
else:
conn.sendlines(request.body, chunked)
if chunked:
conn.send_chunk("")
else:
conn.send(msg)
return self.get_response(request, conn)
except socket.gaierror, e:
if conn is not None:
conn.release(True)
raise RequestError(str(e))
except socket.timeout, e:
if conn is not None:
conn.release(True)
raise RequestTimeout(str(e))
except socket.error, e:
if log.isEnabledFor(logging.DEBUG):
log.debug("socket error: %s" % str(e))
if conn is not None:
conn.close()
if e[0] not in (errno.EAGAIN, errno.EPIPE, errno.EBADF) or \
tries >= self.max_tries:
raise RequestError("socket.error: %s" % str(e))
# should raised an exception in other cases
request.maybe_rewind(msg=str(e))
except BadStatusLine:
if conn is not None:
conn.release(True)
# should raised an exception in other cases
request.maybe_rewind(msg="bad status line")
if tries >= self.max_tries:
raise
except Exception:
# unkown error
log.debug("unhandled exception %s" %
traceback.format_exc())
if conn is not None:
conn.release(True)
raise
tries += 1
self._pool.backend_mod.sleep(self.wait_tries)
def request(self, url, method='GET', body=None, headers=None):
""" perform immediatly a new request """
request = Request(url, method=method, body=body,
headers=headers)
# apply request filters
# They are applied only once time.
for f in self.request_filters:
ret = f.on_request(request)
if isinstance(ret, Response):
# a response instance has been provided.
# just return it. Useful for cache filters
return ret
# no response has been provided, do the request
self._nb_redirections = self.max_follow_redirect
return self.perform(request)
def redirect(self, location, request):
""" reset request, set new url of request and perform it """
if self._nb_redirections <= 0:
raise RedirectLimit("Redirection limit is reached")
if request.initial_url is None:
request.initial_url = self.url
# make sure location follow rfc2616
location = rewrite_location(request.url, location)
if log.isEnabledFor(logging.DEBUG):
log.debug("Redirect to %s" % location)
# change request url and method if needed
request.url = location
self._nb_redirections -= 1
#perform a new request
return self.perform(request)
def get_response(self, request, connection):
""" return final respons, it is only accessible via peform
method """
if log.isEnabledFor(logging.DEBUG):
log.debug("Start to parse response")
p = HttpStream(SocketReader(connection.socket()), kind=1,
decompress=self.decompress)
if log.isEnabledFor(logging.DEBUG):
log.debug("Got response: %s %s" % (p.version(), p.status()))
log.debug("headers: [%s]" % p.headers())
location = p.headers().get('location')
if self.follow_redirect:
if p.status_code() in (301, 302, 307,):
connection.close()
if request.method in ('GET', 'HEAD',) or \
self.force_follow_redirect:
if hasattr(self.body, 'read'):
try:
self.body.seek(0)
except AttributeError:
raise RequestError("Can't redirect %s to %s "
"because body has already been read"
% (self.url, location))
return self.redirect(location, request)
elif p.status_code() == 303 and self.method == "POST":
connection.close()
request.method = "GET"
request.body = None
return self.redirect(location, request)
# create response object
resp = self.response_class(connection, request, p)
# apply response filters
for f in self.response_filters:
f.on_response(resp, request)
if log.isEnabledFor(logging.DEBUG):
log.debug("return response class")
# return final response
return resp
def _get_proxy_auth(proxy_settings):
proxy_username = os.environ.get('proxy-username')
if not proxy_username:
proxy_username = os.environ.get('proxy_username')
proxy_password = os.environ.get('proxy-password')
if not proxy_password:
proxy_password = os.environ.get('proxy_password')
proxy_password = proxy_password or ""
if not proxy_username:
u = urlparse.urlparse(proxy_settings)
if u.username:
proxy_password = u.password or proxy_password
proxy_settings = urlparse.urlunparse((u.scheme,
u.netloc.split("@")[-1], u.path, u.params, u.query,
u.fragment))
if proxy_username:
user_auth = base64.encodestring('%s:%s' % (proxy_username,
proxy_password))
return proxy_settings, 'Basic %s\r\n' % (user_auth.strip())
else:
return proxy_settings, ''