mirai - Promises Integration
In mirai: Minimalist Async Evaluation Framework for R

Promises Integration

mirai supplies its own as.promise() method, allowing it to be used as a promise from the promises package.

These are next-generation, event-driven promises, developed in collaboration with Joe Cheng.

Does not require each promise to be polled for completion by a later loop like other promises.
Instead, promise actions are automatically queued for execution as soon as each 'mirai' resolves.
Allows for much higher responsiveness (lower latency) and massive scalability (thousands or even millions of promises).

A 'mirai' may be piped directly using the promise pipe %...>%, which implicitly calls as.promise() on the 'mirai'. Similarly all promise-aware functions such as promises::then() or shiny::ExtendedTask$new() which take a promise can also take a 'mirai' (using promises >= 1.3.0).

Alternatively, a 'mirai' may be explicitly converted into a promise by as.promise(), which then allows using the methods $then(), $finally() etc.

The following example outputs "hello" to the console after one second when the 'mirai' resolves.

library(mirai)
library(promises)

p <- mirai({Sys.sleep(1); "hello"}) %...>% cat()
p
#> <Promise [pending]>

It is possible to both access a 'mirai' value at $data and to use a promise for enacting a side effect (assigning the value to an environment in the example below).

env <- new.env()

m <- mirai({
  Sys.sleep(1)
  "hello"
})

promises::then(m, function(x) env$res <- x)

m[]
#> [1] "hello"

After returning to the top level prompt:

env$res
#> [1] "hello"

mirai_map Promises Integration

A mirai_map also has an as.promise() method, which allows it to be used directly in a Shiny ExtendedTask.

The implementation uses promises::promise_all(), and will resolve when the entire map operation completes or at least one mirai in the map is rejected.

The following example prints the list 1, 2, 3 to the console after the flatmap returns a vector of the same values.

library(mirai)
library(promises)

with(daemons(2), {
  mp <- mirai_map(1:3, function(x) { Sys.sleep(x); x })
  mp %...>% print
  mp[.flat]
})
#> [1] 1 2 3
#> [[1]]
#> [1] 1
#> 
#> [[2]]
#> [1] 2
#> 
#> [[3]]
#> [1] 3

The One Million Promises Challenge

The code below is taken from the challenge to launch and collect one million promises. For illustration, the example is scaled down to ten thousand.

library(mirai)
daemons(8, dispatcher = FALSE)
#> [1] 8
r <- 0
start <- Sys.time()
m <- mirai_map(1:10000, \(x) x, .promise = \(x) r <<- r + x)
Sys.time() - start
#> Time difference of 3.441591 secs
later::run_now()
r
#> [1] 50005000
daemons(0)
#> [1] 0