DESIGN: Pros and cons of different solutions

[HenrikBengtsson]

Progress updates via '.progress' argument

Example:

y <- future_lapply(..., .progress = function(...) { ... })

Pros:

• Clear that function produces progress info

Cons:

• Requires specifying progress argument
• It is not possible to listen to progress info from nested "internal" functions

Progress updates via centralized "progress" subscription

Example:

progress_subscribe(function(name, step, ...) {
  # called each time there is a progress update
})

y <- future_lapply(...)

Pros:

• Anyone can subscribe
• No "progress" argument

Cons:

• Not clear how to handled nested progress updated
• Not clear who can/should subscribe

Progress updates via condition signaling

Example:

progress_relay({
  y <- future_lapply(...)
})

Pros:

• Anyone can listen from anywhere upstream
• No "progress" argument
• No need to update code when a downstream function introduces progress info (correct?)
• Progress conditions can be captured, filtered, summarized locally in functions and then resignaled in different forms and shapes.

Cons:

• Requires wrapping whole expressions, e.g. relay_progress(y <- foo())
• End user needs to wrap calls to, but we might be able to use a task callback handler to handle progress conditions that bubble up to the top level.