X-Git-Url: https://jfr.im/git/z_archive/twitter.git/blobdiff_plain/d488eec0aa7d61b5d9d49e391ca96ef2b0c78e77..e28a1da9ea6df9761811e101e9e475268a87ac76:/twitter/stream.py

diff --git a/twitter/stream.py b/twitter/stream.py
index 094b5f9..fa332cf 100644
--- a/twitter/stream.py
+++ b/twitter/stream.py
@@ -12,6 +12,28 @@ import sys, select, time
 
 from .api import TwitterCall, wrap_response, TwitterHTTPError
 
+def recv_chunk(sock):
+    buf = sock.recv(10)  # Scan for an up to a 4GiB chunk size (0xffffffff).
+    if buf:
+        crlf = buf.find(b'\r\n')  # Find the HTTP chunk size.
+        if crlf > 0:
+            remaining = int(buf[:crlf], 16)  # Decode the chunk size.
+            chunk = bytearray(remaining)  # Create the chunk buffer.
+
+            start = crlf + 2  # Add in the length of the header's CRLF pair.
+            end = len(buf) - start
+
+            chunk[:end] = buf[start:]
+            chunk[end:] = sock.recv(remaining - end)
+
+            sock.recv(2)  # Read the trailing CRLF pair. Throw it away.
+
+            return chunk
+    return b''
+
+##  recv_chunk()
+
+
 class TwitterJSONIter(object):
 
     def __init__(self, handle, uri, arg_data, block=True, timeout=None):
@@ -19,10 +41,8 @@ class TwitterJSONIter(object):
         self.handle = handle
         self.uri = uri
         self.arg_data = arg_data
-        self.buf = b""
         self.block = block
         self.timeout = timeout
-        self.timer = time.time()
 
 
     def __iter__(self):
@@ -33,36 +53,33 @@ class TwitterJSONIter(object):
         sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
         if not self.block or self.timeout:
             sock.setblocking(False)
+        buf = b''
+        timer = time.time()
         while True:
             try:
-                utf8_buf = self.buf.decode('utf8').lstrip()
-                if utf8_buf and utf8_buf[0] != '{':  # Remove the hex delimiter length and extra whitespace.
-                    utf8_buf = utf8_buf.lstrip('0123456789abcdefABCDEF')
-                    utf8_buf = utf8_buf.lstrip()
+                utf8_buf = buf.decode('utf-8').lstrip()
                 res, ptr = self.decoder.raw_decode(utf8_buf)
-                self.buf = utf8_buf[ptr:].encode('utf8')
+                buf = utf8_buf[ptr:].encode('utf-8')
                 yield wrap_response(res, self.handle.headers)
-                self.timer = time.time()
+                timer = time.time()
                 continue
             except ValueError as e:
                 if self.block:
                     pass
                 else:
                     yield None
-            except urllib_error.HTTPError as e:  # Probably unnecessary, no dynamic url calls in the try block.
-                raise TwitterHTTPError(e, self.uri, 'json', self.arg_data)
             # this is a non-blocking read (ie, it will return if any data is available)
             try:
                 if self.timeout:
                     ready_to_read = select.select([sock], [], [], self.timeout)
                     if ready_to_read[0]:
-                        self.buf += sock.recv(1024)
-                        if time.time() - self.timer > self.timeout:
+                        buf += recv_chunk(sock)
+                        if time.time() - timer > self.timeout:
                             yield {"timeout":True}
                     else:
                         yield {"timeout":True}
                 else:
-                    self.buf += sock.recv(1024)  # As tweets are typically longer than 1KB, consider increasing this size.
+                    buf += recv_chunk(sock)
             except SSLError as e:
                 if (not self.block or self.timeout) and (e.errno == 2):
                     # Apparently this means there was nothing in the socket buf