Print and log before the universe finishes running?

[uwdata]

Print and log before the universe finishes running?

[AmethystGear]

I think we can, but I have a few questions:
Can we create two files, one for stdout and one for stderr in the boba logs?
i.e. for each log we have a folder "log_x", and inside that folder we have stdout.txt and stderr.txt
That would make periodic logging easier, since then stderr doesn't have to wait for stdout to finish before we can start printing it.

Second, what should be the print behavior for multiprocessing?
if we allow printing in the middle of the process we might have something like this happen if we're running multiple jobs at once:

job 1 says hello
job 2 says hello
job 1 says goodbye
job 2 says goodbye

so the printing would be mixed between different running processes. There's no problem with continuous logging since we're logging to different files, but should we disable continuous printing if there are multiple processes running?

[yyyliu]

Good questions :)

• Having two files per universe is perfectly fine. Perhaps we could name stdout as log_x and stderr as error_x, so all logs can still be in the same folder. Maybe we can skip error_x if there is no error.
• Interleaving outputs is not ideal, but it is fine unless we have a better solution ... Is it possible to append something like [universe x] before each line/snippet of the output? If not, maybe we could stick with what we have now, except that we can print something like universe x has started at the beginning so users can get some immediate feedback.

[AmethystGear]

Yes, we can print [universe_x] before each line to tell the user where that line is coming from. I'll make a PR implementing this soon.

[yyyliu]

Great, thanks!