python - ZMQ中的HTTP服务器或如何使用pyzmq处理POST请求？

我正在尝试创建一个带有ZMQ_STREAM套接字的http服务器。
当我做一个简单的POST请求时：

POST  HTTP/1.1
Host: localhost:5555
Cache-Control: no-cache
Postman-Token: 67004be5-56bc-c1a9-847a-7db3195c301d

Apples to Oranges!

下面是我如何处理Pyzmq：

context = zmq.Context()
socket = context.socket(zmq.STREAM)
socket.bind("tcp://*:5555")

while True:
    # Get HTTP request
    parts = []
    id_, msg = socket.recv_multipart()  # [id, ''] or [id, http request]
    parts.append(id_)
    parts.append(msg)
    if not msg:
        # This is a new connection - this is just the identify frame (throw away id_)
        # The body will come next
        id_, msg = socket.recv_multipart() # [id, http request]
        parts.append(id_)
        parts.append(msg)

        end = socket.recv_multipart() # [id*, ''] <- some kind of junk?
        parts.append(end)

    print("%s" % repr(parts))

因此parts列表显示为：

['\x00\x80\x00\x00)', '', '\x00\x80\x00\x00)', 'POST / HTTP/1.1\r\nHost: localhost:5555\r\nConnection: keep-alive\r\nContent-Length: 18\r\nCache-Control: no-cache\r\nOrigin: chrome-extension://fhbjgbiflinjbdggehcddcbncdddomop\r\nContent-Type: text/plain;charset=UTF-8\r\nUser-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/45.0.2454.101 Safari/537.36\r\nPostman-Token: 9503fce9-8b1c-b39c-fb4d-3a7f21b509de\r\nAccept: */*\r\nAccept-Encoding: gzip, deflate\r\nAccept-Language: en-US,en;q=0.8,ru;q=0.6,uk;q=0.4\r\n\r\nApples to Oranges!', ['\x00\x80\x00\x00*', '']]

所以我明白：
'\x00\x80\x00\x00)', ''是连接的标识。最初由ZMQ_STREAM套接字设置。在随后的请求中，它似乎不存在。
\x00\x80\x00\x00)再次是标识，这是我们在从ZMQ_STREAM套接字发出的客户机后续请求中看到的。
那么实际的http请求
但最后一对神奇数字：['\x00\x80\x00\x00*', '']
那是什么意思？
参考文献：
http://api.zeromq.org/4-0:zmq-socket
http 1.1规范：http://www.w3.org/Protocols/rfc2616/rfc2616-sec5.html

最佳答案

但最后一对神奇数字是：['\x00\x80\x00\x00*'，'']
那是什么意思？
这是一个新的连接，有一个新的连接id。连接id是一个整数计数器，您可以看到使用python内置的ord来查看ord(')') = 41和ord('*') = 42，这是序列中的下一个数字。
使用zmq_流编写http服务器时，必须小心，因为它比建立连接后只接收一条消息更复杂。
问题主要是你不能保证一个请求将完成；主体可以在潜在的多个消息中以块的形式到达。您将不得不读取http头并处理以片段形式接收正文的操作。
下面是一个处理来自curl的post请求的示例：

from traceback import print_exc
import zmq
from tornado.httputil import HTTPHeaders

class BadRequest(Exception):
    pass

class ConnectionLost(Exception):
    pass

def parse_request(request):
    """Parse a request verp, path, and headers"""
    first_line, header_lines = request.split(b'\r\n', 1)
    verb, path, proto = first_line.decode('utf8').split()
    headers = HTTPHeaders.parse(header_lines.decode('utf8', 'replace'))
    return verb, path, headers


def recv_body(socket, headers, chunks, request_id):
    """Receive the body of a request"""
    if headers.get('expect', '').lower() == '100-continue':
        if 'Content-Length' not in headers:
            # Don't support chunked transfer: http://tools.ietf.org/html/rfc2616#section-3.6.1
            print("Only support specified-length requests")
            socket.send_multipart([
                request_id, b'HTTP/1.1 400 (Bad Request)\r\n\r\n',
                request_id, b'',
            ])
            msg = 1
            while msg != b'':
                # flush until new connection
                _, msg = socket.recv_multipart()
            raise BadRequest("Only support specified-length requests")

        socket.send_multipart([request_id, b'HTTP/1.1 100 (Continue)\r\n\r\n'], zmq.SNDMORE)

        content_length = int(headers['Content-Length'])
        print("Waiting to receive %ikB body" )
        while sum(len(chunk) for chunk in chunks) < content_length:
            id_, msg = socket.recv_multipart()
            if msg == b'':
                raise ConnectionLost("Disconnected")
            if id_ != request_id:
                raise ConnectionLost("Received data from wrong ID: %s != %s" % (id_, request_id))
            chunks.append(msg)
    return b''.join(chunks)


print(zmq.__version__, zmq.zmq_version())


socket = zmq.Context().socket(zmq.STREAM)
socket.bind("tcp://*:5555")


while True:
    # Get HTTP request
    request_id, msg = socket.recv_multipart()
    if msg == b'':
        continue
    chunks = []
    try:
        request, first_chunk = msg.split(b'\r\n\r\n', 1)
        if first_chunk:
            chunks.append(first_chunk)
        verb, path, headers = parse_request(request)
        print(verb, path)
        print("Headers:")
        for key, value in headers.items():
            print('  %s: %s' % (key, value))
        body = recv_body(socket, headers, chunks, request_id)
        print("Body: %r" % body)
    except BadRequest as e:
        print("Bad Request: %s" % e)
    except ConnectionLost as e:
        print("Connection Lost: %s" % e)
    except Exception:
        print("Failed to handle request", msg)
        print_exc()
        socket.send_multipart([
            request_id, b'HTTP/1.1 500 (OK)\r\n\r\n',
            request_id, b''])
    else:
        socket.send_multipart([
            request_id, b'HTTP/1.1 200 (OK)\r\n\r\n',
            request_id, b''])

这种情况的相关逻辑是在recv_body方法中，它读取头部并继续身体的RCV块直到完成。
坦率地说，我认为用zmq_流用python编写http服务器没有多大意义。您可以将ZMQ套接字与现有Python EngError集成，并重新使用已经建立的HTTP库，这样您就不必再重新发明这个特定的轮子。例如，PyZMQ与龙卷风EntError的关系非常好，可以在同一个应用程序中使用ZMQ套接字和TrnADO-HTTP处理程序。