Problem when implementing scheduler in PyQt5

[giwon2004]

First, I managed to make a scheduler based on the queued experiments fetched from artiq-proxy:

and this shows three experiments waiting to be run.

However, each experiment goes through the entire procedure (after pending) in 0.1s or so.
This means that the scheduler must fetch the information ("status" value in specific, varying among running, analyzing, deleting) almost every 10ms.

And this is what I get using artiq-proxy to fetch the current status value with frequency 1ms:

This monstrous amount of GET requests knocks down the server after a long amount of time, and I believe this can be solved only by modifying the artiq_master.py itself to emit a signal when a data change is made.

Also, it makes repeated (but unnecessary for almost all of the time) refreshing of QListView of the scheduler, along with the problem of making a menu feature as the item gets refreshed too quickly.

I do not think this is the best way to implement it. Would there be any different approach to this problem?

Thank you in advance.

[kangz12345]

As you said, that approach is not feasible.
Did you check how the ARTIQ dashboard scheduler fetches the queue status?

[BECATRUE]

I didn't take a close look at it, in artiq_client.py, the function to show the scheduled list is implemented using Subscriber.
It may help you.

[giwon2004]

In artiq/artiq/frontend/artiq_client.py we have _show_schedule which displays a dictionary schedule containing dictionaries of experiments, and this schedule is repeatedly refreshed by Subscriber through _show_dict. I guess I can use Subscriber with its parameter notify_cb which is a perfect one I need to use.

[kangz12345]

Oh, I found "HTTP long polling".
In summary,

• the client sends a request and wait for a long time (poll and wait)
• the server receives the request and respond when the requested response is ready

Since our proxy server is asynchronous, I think we can implement this concept quite easily.

[giwon2004]

No, GUI did not freeze.

[giwon2004]

How do I suspend a thread?

global current_queue, previous_queue
    remote = get_client("master_schedule")
    while current_queue == previous_queue:
        time.sleep(0.01)
        current_queue = remote.get_status()
    previous_queue = current_queue
    return current_queue

seems failing.

[kangz12345]

Since you are using asyncio (async def ...), you can use await asyncio.sleep(0) to suspend the current task.

sleep() always suspends the current task, allowing other tasks to run.
Setting the delay to 0 provides an optimized path to allow other tasks to run. This can be used by long-running functions to avoid blocking the event loop for the full duration of the function call.

[giwon2004]

It worked!! Thank you for your advice.

[kangz12345]

Oh that's great!

[giwon2004]

If I use a Subscriber instance to implement this feature, then there will be no change on main.py in artiq-proxy, as Subscriber can automatically call the refreshing function (via notify_cb) which would be enough calling the refreshing function _snycExperimentQueue introduced in my branch.