Signal mechanism from daemon to host

[shikokuchuo]

(This message has been a journey.)

After more research, I cannot find a specific "signal" mechanism so that the R expression can immediately be signaled back to the main host while the rest of the expression continues to execute.

Long-story-short, I enabled daemons(..., output=TRUE) to see errors/warnings immediately, nice that they are passed immediately. I thought they may be passed back to the local process as "condition"s in which case handlers would be able to work on them, but alas I believe they are being merely cat'ed (or equivalent) to the local console.

From this, two questions:

• Is it possible to tag the remote-output with a particular "condition" where its default local handler is merely cat(cond)? This would continue the current behavior, but would allow a calling handler to deal with remote output distinctly from local output. (If always doing this as a condition is concerning, it could be as an option, something like daemons(.., output=T, output_condition=T).)
• I see the since-removed .signal= argument which references nanonext::cv(), is cv() applicable here? Is there a way to allow remote signals (distinct from console output) to be caught via the normal R mechanisms?

I'm adding this comment mostly to say that I continue to look into the source to hope to self-answer, but I'm not finding something to lead me to a working solution. Since this discussion has spiraled from unassigned() to a bit more, would it be better as a Discussion?

Originally posted by @r2evans in #540

[shikokuchuo]

Some concrete motivation for this would be helpful, i.e. why do you need the daemon to send back signals vs. breaking up the expression and just returning values each time.

(signalling condition variables aren't an answer as they are purely internal, and also intra- not inter-process).

[r2evans]

The motivation is to determine starting conditions for the newly-assigned task. For now this is just "time", but there's no reason it could not include other attributes. In my head it's mostly meta, not "data" or intermediate results.