Feature Request: Option to automatically run revise in a parallel task

[schlichtanders]

Hello,
I am using Revise with a Webserver and found this discourse thread with the hint that one needs to execute something on the REPL in order to trigger revise().

It would be great if Revise.jl had a mode where this update is done automatically without the need of another execution in the REPL. This would simplify the workflow massively when working with an ongoing process like a webserver.

[KristofferC]

FWIW, here is how it is done in another webserver:

https://github.com/JuliaComputing/FemtoCleaner.jl/blob/384815c983609275ae8bdc0aa0dcc67a287d1768/src/FemtoCleaner.jl#L507-L514

[schlichtanders]

adding an explicit revise() call to the webserver is a workaround, but requires the Webserver package to depend on Revise (even package extensions don't work, as this would involve type piracy against its own main package, which recently got raised to error status in julia 1.10).

Hence, it would be great if the auto updates could be done by Revise itself in an async task.

[timholy]

What event would trigger revise in this async task? Would it just be on a timer? The concern I have is that you still probably want to trigger Revise at "appropriate" times. E.g., adding a method

foo(::NewType) = nothing

to src/code.jl before adding

struct NewType end

to src/types.jl is obviously problematic. But developers like me do this all the time (it's a good opportunity to play with your struct before actually defining it), and because Revise currently gets triggered by trying a command at the REPL, it only errors if the developer has genuinely made a mistake, and it probably only happens once. But if it auto-triggers, then perfectly valid developer practices will start throwing errors, and those errors may repeat every second if it's on a timer. This would be extremely annoying.

I recognize that this may perhaps not be what you're asking for, in which case a bit more detail about the request would be very helpful.