ipython / ipython/ipykernel

[Question] Asynchronous Comm.on_msg callback

未關閉
#542 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視
主要語言
Python
星號
734
分支
412
平均合併
1 天 5 小時
30 天內合併 PR
8

描述

I'm registering a callback function that forward every messages associated to given registered target to a ZMQ socket (which is a remote server). It works great when the main thread is idle, but if the main thread is busy for some reason, the callback is not call until the main thread finishes its processing. It is a big issue in by case because in practice I'm busy waiting for the server to acknowledge it has received the data...

Here is how I'm registering the callback:
```
class ZMQIPythonBridge(threading.Thread):
def __init__(self, comm_url):
threading.Thread.__init__(self)
self.zmq_url = zmq_url

def comm_register(self, comm, msg):
@comm.on_msg
def _on_msg(msg):
data = msg['content']['data']
self.comm_zmq.send(f"data:{comm.comm_id}:{data}".encode())

self.comm_zmq.send(f"open:{comm.comm_id}".encode())

def run(self):
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
ioloop = tornado.ioloop.IOLoop()
ioloop.make_current()
context = zmq.Context()
self.zmq_socket = context.socket(zmq.XREQ)
self.zmq_socket.connect(self.zmq_url)
get_ipython().kernel.comm_manager.register_target(
'my_channel', self.comm_register)
ioloop.start()

comm_bridge = ZMQIPythonBridge(comm_url)
comm_bridge.start()
```

Is there a way to still be enable to trigger `_on_msg` while the main thread is busy wait for the ZMQ socket to acknowledge ? I tried to use a different thread, but as expected in it not changing anything (since the callback mechanism is still handle by the main thread...)

貢獻指南

開啟貢獻指南

研究方向

Start by tracing the `comm.on_msg` callback, `comm_manager.register_target`, and the Tornado `IOLoop` setup shown in the issue, then reproduce the behavior while the main thread busy-waits on the ZMQ socket. Done means determining whether callbacks can run during that wait and documenting the supported approach or limitation.

由索引模型根據 Issue 內容生成。

評估

技術堆疊
jupyter, python
領域
api, backend
Issue 類型
缺陷
難度
4/5
預估耗時
3-5 天
活躍度
停滯
描述清晰度
需要釐清
新手友好度
25/100

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。