X-Git-Url: https://jfr.im/git/z_archive/twitter.git/blobdiff_plain/b8fd12068076e2be6bfb44ee1ff64c6f8fd8b497..443e409dca0f1808eb10692e14d1a666d0a03f66:/twitter/stream.py diff --git a/twitter/stream.py b/twitter/stream.py index 2c60918..96e92d7 100644 --- a/twitter/stream.py +++ b/twitter/stream.py @@ -6,42 +6,142 @@ except ImportError: import urllib2 as urllib_request import urllib2 as urllib_error import json +from ssl import SSLError +import socket +import sys, select, time + +from .api import TwitterCall, wrap_response, TwitterHTTPError + +def recv_chunk(sock): # -> bytearray: + + buf = sock.recv(8) # Scan for an up to 16MiB chunk size (0xffffff). + crlf = buf.find(b'\r\n') # Find the HTTP chunk size. + + if crlf > 0: # If there is a length, then process it + + remaining = int(buf[:crlf], 16) # Decode the chunk size. + + start = crlf + 2 # Add in the length of the header's CRLF pair. + end = len(buf) - start + + chunk = bytearray(remaining) + + if end < remaining: + chunk[:end] = buf[start:] + chunk[end:] = sock.recv(remaining - end) + sock.recv(2) # Read the trailing CRLF pair. Throw it away. + else: # E.g. an HTTP chunk with just a keep-alive delimiter. + chunk[:remaining] = buf[start:start + remaining] + + return chunk + + return bytearray() + +## recv_chunk() -from .api import TwitterCall, wrap_response class TwitterJSONIter(object): - def __init__(self, handle, uri, arg_data): - self.decoder = json.JSONDecoder() + def __init__(self, handle, uri, arg_data, block=True, timeout=None): self.handle = handle - self.buf = b"" + self.uri = uri + self.arg_data = arg_data + self.block = block + self.timeout = timeout + def __iter__(self): + sock = self.handle.fp.raw._sock if sys.version_info >= (3, 0) else self.handle.fp._sock.fp._sock + sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1) + sock.setblocking(self.block and not self.timeout) + buf = u'' + json_decoder = json.JSONDecoder() + timer = time.time() while True: - self.buf += self.handle.read(1024) try: - utf8_buf = self.buf.decode('utf8').lstrip() - res, ptr = self.decoder.raw_decode(utf8_buf) - self.buf = utf8_buf[ptr:].encode('utf8') + buf = buf.lstrip() + res, ptr = json_decoder.raw_decode(buf) + buf = buf[ptr:] yield wrap_response(res, self.handle.headers) - except ValueError as e: + timer = time.time() continue - except urllib_error.HTTPError as e: - raise TwitterHTTPError(e, uri, self.format, arg_data) + except ValueError as e: + if self.block: pass + else: yield None + try: + buf = buf.lstrip() # Remove any keep-alive delimiters to detect hangups. + if self.timeout: + ready_to_read = select.select([sock], [], [], self.timeout) + if ready_to_read[0]: + buf += recv_chunk(sock).decode('utf-8') # This is a non-blocking read. + if time.time() - timer > self.timeout: + yield {'timeout': True} + else: + yield {'timeout': True} + else: + buf += recv_chunk(sock).decode('utf-8') + if not buf and self.block: + yield {'hangup': True} + except SSLError as e: + # Error from a non-blocking read of an empty buffer. + if (not self.block or self.timeout) and (e.errno == 2): pass + else: raise -class TwitterStreamCall(TwitterCall): - def _handle_response(self, req, uri, arg_data): +def handle_stream_response(req, uri, arg_data, block, timeout=None): + try: handle = urllib_request.urlopen(req,) - return iter(TwitterJSONIter(handle, uri, arg_data)) + except urllib_error.HTTPError as e: + raise TwitterHTTPError(e, uri, 'json', arg_data) + return iter(TwitterJSONIter(handle, uri, arg_data, block, timeout=timeout)) + +class TwitterStreamCallWithTimeout(TwitterCall): + def _handle_response(self, req, uri, arg_data, _timeout=None): + return handle_stream_response(req, uri, arg_data, block=True, timeout=self.timeout) + +class TwitterStreamCall(TwitterCall): + def _handle_response(self, req, uri, arg_data, _timeout=None): + return handle_stream_response(req, uri, arg_data, block=True) + +class TwitterStreamCallNonBlocking(TwitterCall): + def _handle_response(self, req, uri, arg_data, _timeout=None): + return handle_stream_response(req, uri, arg_data, block=False) class TwitterStream(TwitterStreamCall): + """ + The TwitterStream object is an interface to the Twitter Stream API + (stream.twitter.com). This can be used pretty much the same as the + Twitter class except the result of calling a method will be an + iterator that yields objects decoded from the stream. For + example:: + + twitter_stream = TwitterStream(auth=OAuth(...)) + iterator = twitter_stream.statuses.sample() + + for tweet in iterator: + ...do something with this tweet... + + The iterator will yield tweets forever and ever (until the stream + breaks at which point it raises a TwitterHTTPError.) + + The `block` parameter controls if the stream is blocking. Default + is blocking (True). When set to False, the iterator will + occasionally yield None when there is no available message. + """ def __init__( - self, domain="stream.twitter.com", secure=False, auth=None, - api_version='1'): + self, domain="stream.twitter.com", secure=True, auth=None, + api_version='1.1', block=True, timeout=None): uriparts = () uriparts += (str(api_version),) + if block: + if timeout: + call_cls = TwitterStreamCallWithTimeout + else: + call_cls = TwitterStreamCall + else: + call_cls = TwitterStreamCallNonBlocking + TwitterStreamCall.__init__( self, auth=auth, format="json", domain=domain, - callable_cls=TwitterStreamCall, - secure=secure, uriparts=uriparts) + callable_cls=call_cls, + secure=secure, uriparts=uriparts, timeout=timeout, gzip=False)