X-Git-Url: https://jfr.im/git/z_archive/twitter.git/blobdiff_plain/30af6c0f49cd4a11409f0ff1d78dab696e9c7749..c0c66ba0ebe25c1ddc446111dc29a1e6fc07b367:/twitter/stream.py diff --git a/twitter/stream.py b/twitter/stream.py index 4b3901b..5384d5b 100644 --- a/twitter/stream.py +++ b/twitter/stream.py @@ -1,3 +1,6 @@ +# encoding: utf-8 +from __future__ import unicode_literals + import sys PY_3_OR_HIGHER = sys.version_info >= (3, 0) @@ -10,39 +13,45 @@ else: import json from ssl import SSLError import socket -import io import codecs import sys, select, time from .api import TwitterCall, wrap_response, TwitterHTTPError CRLF = b'\r\n' +MIN_SOCK_TIMEOUT = 0.0 # Apparenty select with zero wait is okay! +MAX_SOCK_TIMEOUT = 10.0 +HEARTBEAT_TIMEOUT = 90.0 Timeout = {'timeout': True} Hangup = {'hangup': True} -HeartbeatTimeout = {'heartbeat_timeout': True, 'hangup': True} - -class ChunkDecodeError(Exception): - pass - -class EndOfStream(Exception): - pass - -range = range if PY_3_OR_HIGHER else xrange +DecodeError = {'hangup': True, 'decode_error': True} +HeartbeatTimeout = {'hangup': True, 'heartbeat_timeout': True} -class HttpDeChunker(object): +class HttpChunkDecoder(object): def __init__(self): self.buf = bytearray() + self.munch_crlf = False - def extend(self, data): - self.buf.extend(data) - - def read_chunks(self): # -> [bytearray] + def decode(self, data): # -> (bytearray, end_of_stream, decode_error) chunks = [] buf = self.buf + munch_crlf = self.munch_crlf + end_of_stream = False + decode_error = False + buf.extend(data) while True: + if munch_crlf: + # Dang, Twitter, you crazy. Twitter only sends a terminating + # CRLF at the beginning of the *next* message. + if len(buf) >= 2: + buf = buf[2:] + munch_crlf = False + else: + break + header_end_pos = buf.find(CRLF) if header_end_pos == -1: break @@ -52,34 +61,35 @@ class HttpDeChunker(object): try: chunk_len = int(header.decode('ascii'), 16) except ValueError: - raise ChunkDecodeError() + decode_error = True + break if chunk_len == 0: - raise EndOfStream() + end_of_stream = True + break data_end_pos = data_start_pos + chunk_len - if len(buf) > data_end_pos + 2: + if len(buf) >= data_end_pos: chunks.append(buf[data_start_pos:data_end_pos]) - buf = buf[data_end_pos + 2:] + buf = buf[data_end_pos:] + munch_crlf = True else: break self.buf = buf - return chunks + self.munch_crlf = munch_crlf + return bytearray().join(chunks), end_of_stream, decode_error -class JsonDeChunker(object): +class JsonDecoder(object): def __init__(self): - self.buf = u"" + self.buf = "" self.raw_decode = json.JSONDecoder().raw_decode - def extend(self, data): - self.buf += data - - def read_json_chunks(self): + def decode(self, data): chunks = [] - buf = self.buf + buf = self.buf + data while True: try: buf = buf.lstrip() @@ -93,6 +103,7 @@ class JsonDeChunker(object): class Timer(object): + def __init__(self, timeout): # If timeout is None, we never expire. self.timeout = timeout @@ -113,69 +124,84 @@ class Timer(object): return False +class SockReader(object): + def __init__(self, sock, sock_timeout): + self.sock = sock + self.sock_timeout = sock_timeout + + def read(self): + try: + ready_to_read = select.select([self.sock], [], [], self.sock_timeout)[0] + if ready_to_read: + return self.sock.read() + except SSLError as e: + # Code 2 is error from a non-blocking read of an empty buffer. + if e.errno != 2: + raise + return bytearray() + + class TwitterJSONIter(object): def __init__(self, handle, uri, arg_data, block, timeout, heartbeat_timeout): self.handle = handle self.uri = uri self.arg_data = arg_data - self.block = block - self.timeout = float(timeout) if timeout else None - self.heartbeat_timeout = float(heartbeat_timeout) if heartbeat_timeout else None - + self.timeout_token = Timeout + self.timeout = None + self.heartbeat_timeout = HEARTBEAT_TIMEOUT + if timeout and timeout > 0: + self.timeout = float(timeout) + elif not (block or timeout): + self.timeout_token = None + self.timeout = MIN_SOCK_TIMEOUT + if heartbeat_timeout and heartbeat_timeout > 0: + self.heartbeat_timeout = float(heartbeat_timeout) def __iter__(self): - actually_block = self.block and not self.timeout - sock_timeout = min(self.timeout or 1000000, self.heartbeat_timeout) + timeouts = [t for t in (self.timeout, self.heartbeat_timeout, MAX_SOCK_TIMEOUT) + if t is not None] + sock_timeout = min(*timeouts) sock = self.handle.fp.raw._sock if PY_3_OR_HIGHER else self.handle.fp._sock.fp._sock sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1) - sock.setblocking(actually_block) headers = self.handle.headers - dechunker = HttpDeChunker() - utf8decoder = codecs.getincrementaldecoder("utf-8")() - json_dechunker = JsonDeChunker() + sock_reader = SockReader(sock, sock_timeout) + chunk_decoder = HttpChunkDecoder() + utf8_decoder = codecs.getincrementaldecoder("utf-8")() + json_decoder = JsonDecoder() timer = Timer(self.timeout) heartbeat_timer = Timer(self.heartbeat_timeout) + while True: - json_chunks = json_dechunker.read_json_chunks() - for json in json_chunks: - yield wrap_response(json, headers) - if json_chunks: - timer.reset() + # Decode all the things: + data = sock_reader.read() + dechunked_data, end_of_stream, decode_error = chunk_decoder.decode(data) + unicode_data = utf8_decoder.decode(dechunked_data) + json_data = json_decoder.decode(unicode_data) + + # Yield data-like things: + for json_obj in json_data: + yield wrap_response(json_obj, headers) + + # Reset timers: + if dechunked_data: heartbeat_timer.reset() + if json_data: + timer.reset() - if not self.block and not self.timeout: - yield None + # Yield timeouts and special things: + if end_of_stream: + yield Hangup + break + if decode_error: + yield DecodeError + break if heartbeat_timer.expired(): yield HeartbeatTimeout break if timer.expired(): - yield Timeout + yield self.timeout_token - try: - ready_to_read = select.select([sock], [], [], sock_timeout)[0] - if not ready_to_read: - continue - data = sock.read() - except SSLError as e: - # Code 2 is error from a non-blocking read of an empty buffer. - if e.errno != 2: - raise - continue - - dechunker.extend(data) - - try: - chunks = dechunker.read_chunks() - except (ChunkDecodeError, EndOfStream): - yield Hangup - break - - for chunk in chunks: - if chunk: - json_dechunker.extend(utf8decoder.decode(chunk)) - if chunks: - heartbeat_timer.reset() def handle_stream_response(req, uri, arg_data, block, timeout, heartbeat_timeout): try: @@ -195,7 +221,37 @@ class TwitterStream(TwitterCall): iterator = twitter_stream.statuses.sample() for tweet in iterator: - ...do something with this tweet... + # ...do something with this tweet... + + Per default the ``TwitterStream`` object uses + [public streams](https://dev.twitter.com/docs/streaming-apis/streams/public). + If you want to use one of the other + [streaming APIs](https://dev.twitter.com/docs/streaming-apis), specify the URL + manually: + + - [Public streams](https://dev.twitter.com/docs/streaming-apis/streams/public): stream.twitter.com + - [User streams](https://dev.twitter.com/docs/streaming-apis/streams/user): userstream.twitter.com + - [Site streams](https://dev.twitter.com/docs/streaming-apis/streams/site): sitestream.twitter.com + + Note that you require the proper + [permissions](https://dev.twitter.com/docs/application-permission-model) to + access these streams. E.g. for direct messages your + [application](https://dev.twitter.com/apps) needs the "Read, Write & Direct + Messages" permission. + + The following example demonstrates how to retrieve all new direct messages + from the user stream:: + + auth = OAuth( + consumer_key='[your consumer key]', + consumer_secret='[your consumer secret]', + token='[your token]', + token_secret='[your token secret]' + ) + twitter_userstream = TwitterStream(auth=auth, domain='userstream.twitter.com') + for msg in twitter_userstream.user(): + if 'direct_message' in msg: + print msg['direct_message']['text'] The iterator will yield until the TCP connection breaks. When the connection breaks, the iterator yields `{'hangup': True}`, and