1 import sys, os, time, md5, random
2 from mod_python import apache, util
3 import osrf.system, osrf.cache, osrf.json, osrf.conf, osrf.net, osrf.log
4 from osrf.const import OSRF_MESSAGE_TYPE_DISCONNECT, OSRF_MESSAGE_TYPE_CONNECT, \
5 OSRF_STATUS_CONTINUE, OSRF_STATUS_TIMEOUT, OSRF_MESSAGE_TYPE_STATUS
9 Proof of concept OpenSRF-HTTP multipart streaming gateway.
11 Example Apache mod_python config:
13 <Location /osrf-http-translator>
15 PythonPath "['/path/to/osrf-python'] + sys.path"
16 PythonHandler osrf.http_translator
17 PythonOption OSRF_CONFIG /path/to/opensrf_core.xml
18 PythonOption OSRF_CONFIG_CONTEXT config.gateway
19 PythonOption OSRF_CACHE_SERVERS 127.0.0.1:11211
25 OSRF_HTTP_HEADER_TO = 'X-OpenSRF-to'
26 OSRF_HTTP_HEADER_XID = 'X-OpenSRF-thread'
27 OSRF_HTTP_HEADER_FROM = 'X-OpenSRF-from'
28 OSRF_HTTP_HEADER_THREAD = 'X-OpenSRF-thread'
29 OSRF_HTTP_HEADER_TIMEOUT = 'X-OpenSRF-timeout'
30 OSRF_HTTP_HEADER_SERVICE = 'X-OpenSRF-service'
31 OSRF_HTTP_HEADER_MULTIPART = 'X-OpenSRF-multipart'
32 MULTIPART_CONTENT_TYPE = 'multipart/x-mixed-replace;boundary="%s"'
33 JSON_CONTENT_TYPE = 'text/plain'
39 # If DEBUG_WRITE = True, all data sent to the client is also written
40 # to stderr (apache error log)
45 sys.stderr.write("%s\n\n" % str(msg))
51 ''' At time of writing, mod_python doesn't support a child_init handler,
52 so this function is called once per process to initialize
53 the opensrf connection '''
55 global INIT_COMPLETE, ROUTER_NAME, OSRF_DOMAIN
59 # Apache complains with: UnboundLocalError: local variable 'osrf' referenced before assignment
60 # if the following import line is removed, even though its also at the top of the file...
63 ops = req.get_options()
64 conf = ops['OSRF_CONFIG']
65 ctxt = ops.get('OSRF_CONFIG_CONTEXT') or 'opensrf'
66 osrf.system.System.net_connect(config_file=conf, config_context=ctxt)
68 ROUTER_NAME = osrf.conf.get('router_name')
69 OSRF_DOMAIN = osrf.conf.get('domain')
72 servers = ops.get('OSRF_CACHE_SERVERS')
74 servers = servers.split(',')
76 # no cache servers configured, see if we can talk to the settings server
78 servers = osrf.set.get('cache.global.servers.server')
79 if not isinstance(servers, list):
82 osrf.cache.CacheClient.connect(servers)
86 ''' Create the translator and tell it to process the request. '''
88 status = HTTPTranslator(req).process()
89 osrf.log.log_debug("translator call resulted in status %d" % int(status))
92 class HTTPTranslator(object):
93 def __init__(self, apreq):
100 post = util.parse_qsl(apreq.read(int(apreq.headers_in['Content-length'])))
101 self.body = [d for d in post if d[0] == 'osrf-msg'][0][1]
107 self.complete = False
108 self.handle = osrf.net.get_network_handle()
109 self.handle.set_receive_callback(None)
111 self.recipient = apreq.headers_in.get(OSRF_HTTP_HEADER_TO)
112 self.service = apreq.headers_in.get(OSRF_HTTP_HEADER_SERVICE)
113 self.thread = apreq.headers_in.get(OSRF_HTTP_HEADER_THREAD) or \
114 "%s%s" % (os.getpid(), time.time())
115 self.timeout = apreq.headers_in.get(OSRF_HTTP_HEADER_TIMEOUT) or 1200
116 self.multipart = str(
117 apreq.headers_in.get(OSRF_HTTP_HEADER_MULTIPART)).lower() == 'true'
118 self.connect_only = False
119 self.disconnect_only = False
121 # generate a random multipart delimiter
123 mpart.update("%f%d%d" % (time.time(), os.getpid(), \
124 random.randint(100, 10000000)))
125 self.delim = mpart.hexdigest()
126 self.remote_host = self.apreq.get_remote_host(apache.REMOTE_NOLOOKUP)
127 self.cache = osrf.cache.CacheClient()
132 if self.apreq.header_only:
135 return apache.HTTP_BAD_REQUEST
136 if not self.set_to_addr():
137 return apache.HTTP_BAD_REQUEST
138 if not self.parse_request():
139 return apache.HTTP_BAD_REQUEST
141 while self.handle.recv(0):
142 pass # drop stale messages
145 net_msg = osrf.net.NetworkMessage(
146 recipient=self.recipient, thread=self.thread, body=self.body)
147 self.handle.send(net_msg)
149 if self.disconnect_only:
150 osrf.log.log_debug("exiting early on DISCONNECT")
154 while not self.complete:
156 net_msg = self.handle.recv(self.timeout)
158 return apache.GATEWAY_TIME_OUT
160 if not self.check_status(net_msg):
164 self.init_headers(net_msg)
168 self.respond_chunk(net_msg)
169 if self.connect_only:
172 self.messages.append(net_msg.body)
174 # condense the sets of arrays into a single array of messages
175 if self.complete or self.connect_only:
176 json = self.messages.pop(0)
177 while len(self.messages) > 0:
178 msg = self.messages.pop(0)
179 json = "%s,%s" % (json[0:len(json)-1], msg[1:])
181 self.write("%s" % json)
186 def parse_request(self):
188 If this is solely a DISCONNECT message, we set self.disconnect_only
190 @return True if the body parses correctly, False otherwise
192 osrf_msgs = osrf.json.to_object(self.body)
196 if len(osrf_msgs) == 1:
197 if osrf_msgs[0].type() == OSRF_MESSAGE_TYPE_CONNECT:
198 self.connect_only = True
199 elif osrf_msgs[0].type() == OSRF_MESSAGE_TYPE_DISCONNECT:
200 self.disconnect_only = True
205 def set_to_addr(self):
206 ''' Determines the TO address. Returns false if
207 the address is missing or ambiguous.
208 Also returns false if an explicit TO is specified and the
209 thread/IP/TO combination is not found in the session cache
213 osrf.log.log_warn("specifying both SERVICE and TO is not allowed")
215 self.recipient = "%s@%s/%s" % \
216 (ROUTER_NAME, OSRF_DOMAIN, self.service)
217 osrf.log.log_debug("set service to %s" % self.recipient)
221 # If the client specifies a specific TO address, verify it's
222 # the same address that was cached with the previous request.
223 obj = self.cache.get(self.thread)
224 if obj and obj['ip'] == self.remote_host and \
225 obj['jid'] == self.recipient:
227 osrf.log.log_warn("client [%s] attempted to send directly "
228 "[%s] without a session" % (self.remote_host, self.recipient))
232 def init_headers(self, net_msg):
233 osrf.log.log_debug("initializing request headers")
234 self.apreq.headers_out[OSRF_HTTP_HEADER_FROM] = net_msg.sender
235 self.apreq.headers_out[OSRF_HTTP_HEADER_THREAD] = self.thread
237 self.apreq.content_type = MULTIPART_CONTENT_TYPE % self.delim
238 self.write("--%s\n" % self.delim)
240 self.apreq.content_type = JSON_CONTENT_TYPE
241 self.cache.put(self.thread, \
242 {'ip':self.remote_host, 'jid': net_msg.sender}, CACHE_TIME)
244 osrf.log.log_debug("caching session [%s] for host [%s] and server "
245 " drone [%s]" % (self.thread, self.remote_host, net_msg.sender))
249 def check_status(self, net_msg):
250 ''' Checks the status of the server response.
251 If we received a timeout message, we drop it.
252 if it's any other non-continue status, we mark this session as
253 complete and carry on.
254 @return False if there is no data to return to the caller
255 (dropped message, eg. timeout), True otherwise '''
257 osrf.log.log_debug('checking status...')
258 osrf_msgs = osrf.json.to_object(net_msg.body)
259 last_msg = osrf_msgs.pop()
261 if last_msg.type() == OSRF_MESSAGE_TYPE_STATUS:
262 code = int(last_msg.payload().statusCode())
263 osrf.log.log_debug("got a status message with code %d" % code)
265 if code == OSRF_STATUS_TIMEOUT:
266 osrf.log.log_debug("removing cached session [%s] and "
267 "dropping TIMEOUT message" % net_msg.thread)
268 self.cache.delete(net_msg.thread)
271 if code != OSRF_STATUS_CONTINUE:
277 def respond_chunk(self, resp):
278 ''' Writes a single multipart-delimited chunk of data '''
280 self.write("Content-type: %s\n\n" % JSON_CONTENT_TYPE)
281 self.write("%s\n\n" % resp.body)
283 self.write("--%s--\n" % self.delim)
285 self.write("--%s\n" % self.delim)
288 def write(self, msg):
289 ''' Writes data to the client stream. '''
292 sys.stderr.write(msg)
294 self.apreq.write(msg)