simonsobs / nextline-graphql Goto Github PK

• allow users to select a script to execute from a list of scripts

Sometimes, the GraphQL Subscriptions don't work from Firefox. It is not always reproducible. It might be that the ws works but wss doesn't. Need to investigate.

• store information about each execution, e.g., timestamp, script revisions

I find that print statements don't consistently show in the web page stdout box if I run the script using the "continue" option. They do show if I step through the program line by line with "next" though.

This can be demonstrated by executing:

print('hello world')

with both the next and continue buttons.

Perhaps more dramatically with a loop of prints:

for i in range(100):
    print(i)

In this case I sometimes get 0-3 lines, but the remaining ~97 don't print. Expected output does show in the docker logs on the backend.

I tried to switch from the Python standard logging to loguru. But I will need to do more work to complete it. I am working on the branch loguru.

The loguru readme shows how to send logging messages to loguru sinks. In nextline, this method doesn't correctly show the location of logging messages created in spawned processes. The locations all become nextline.utils.multiprocessing_logging:95, which is where nextline collects logging messages created in spawned processes.

loguru doesn't read the locations from the log records. It determines the locations from the frames in the main process.

On the other hand, loguru does support multiprocessing.. But it doesn't collect logging messages in the main process. It processes logging messages in each sub-process.

Consequently, to use loguru in nextline, we need to disable MultiprocessingLogging and use InterceptHandler in each spawned process.