Package 'doFuture'

Description

Loop over a Foreach Expression using Futures

Usage

foreach %dofuture% expr

Arguments

Details

This is a replacement for %dopar% of the foreach package that leverages the future framework.

When using ⁠%dofuture%⁠:

• there is no need to use registerDoFuture()

• there is no need to use ⁠%dorng%⁠ of the doRNG package (but you need to specify .options.future = list(seed = TRUE) whenever using random numbers in the expr expression)

• global variables and packages are identified automatically by the future framework

• errors are relayed as-is (with ⁠%dopar%⁠ they captured and modified)

Value

The value of the foreach call.

Global variables and packages

When using ⁠%dofuture%⁠, the future framework identifies globals and packages automatically (via static code inspection). However, there are cases where it fails to find some of the globals or packages. When this happens, one can specify the future::future() arguments globals and packages via foreach argument .options.future. For example, if you specify argument .options.future = list(globals = structure(TRUE, ignore = "b", add = "a")) then globals are automatically identified (TRUE), but it ignores b and always adds a.

An alternative to specifying the globals and the packages options via .options.future, is to use the %globals% and the %packages% operators. See the examples for an illustration.

For further details and instructions, see future::future().

Random Number Generation (RNG)

The ⁠%dofuture%⁠ uses the future ecosystem to generate proper random numbers in parallel in the same way they are generated in, for instance, future.apply. For this to work, you need to specify .options.future = list(seed = TRUE). For example,

y <- foreach(i = 1:3, .options.future = list(seed = TRUE)) %dofuture% {
  rnorm(1)
}

Unless seed is FALSE or NULL, this guarantees that the exact same sequence of random numbers are generated given the same initial seed / RNG state - this regardless of type of future backend, number of workers, and scheduling ("chunking") strategy.

RNG reproducibility is achieved by pregenerating the random seeds for all iterations by using L'Ecuyer-CMRG RNG streams. In each iteration, these seeds are set before evaluating the foreach expression. Note, for large number of iterations this may introduce a large overhead.

If seed = TRUE, then .Random.seed is used if it holds a L'Ecuyer-CMRG RNG seed, otherwise one is created randomly.

If seed = FALSE, it is expected that none of the foreach iterations use random number generation. If they do, then an informative warning or error is produces depending on settings. See future::future for more details. Using seed = NULL, is like seed = FALSE but without the check whether random numbers were generated or not.

As input, seed may also take a fixed initial seed (integer), either as a full L'Ecuyer-CMRG RNG seed (vector of 1+6 integers), or as a seed generating such a full L'Ecuyer-CMRG seed. This seed will be used to generated one L'Ecuyer-CMRG RNG stream for each iteration.

An alternative to specifying the seed option via .options.future, is to use the %seed% operator. See the examples for an illustration.

For further details and instructions, see future.apply::future_lapply().

Load balancing ("chunking")

Whether load balancing ("chunking") should take place or not can be controlled by specifying either argument ⁠.options.future = list(scheduling = <ratio>)⁠ or ⁠.options.future = list(chunk.size = <count>)⁠ to foreach().

The value chunk.size specifies the average number of elements processed per future ("chunks"). If +Inf, then all elements are processed in a single future (one worker). If NULL, then argument future.scheduling is used.

The value scheduling specifies the average number of futures ("chunks") that each worker processes. If 0.0, then a single future is used to process all iterations; none of the other workers are not used. If 1.0 or TRUE, then one future per worker is used. If 2.0, then each worker will process two futures (if there are enough iterations). If +Inf or FALSE, then one future per iteration is used. The default value is scheduling = 1.0.

For further details and instructions, see future.apply::future_lapply().

Control processing order of iterations

Attribute ordering of chunk.size or scheduling can be used to control the ordering the elements are iterated over, which only affects the processing order and not the order values are returned. This attribute can take the following values:

• index vector - an numeric vector of length nX.

• function - an function taking one argument which is called as ordering(nX) and which must return an index vector of length nX, e.g. function(n) rev(seq_len(n)) for reverse ordering.

• "random" - this will randomize the ordering via random index vector sample.int(nX).

where nX is the number of foreach iterations to be done.

For example, .options.future = list(scheduling = structure(2.0, ordering = "random")).

Note, when elements are processed out of order, then captured standard output and conditions are also relayed in that order, that is, out of order.

For further details and instructions, see future.apply::future_lapply().

Reporting on progress

How to report on progress is a frequently asked question, especially in long-running tasks and parallel processing. The foreach framework does not have a built-in mechanism for progress reporting(*).

When using doFuture, and the Futureverse in general, for processing, the progressr package can be used to signal progress updates in a near-live fashion. There is special argument related to foreach() or doFuture to achieve this. Instead, one calls a a, so called, "progressor" function within each iteration. See the progressr package and its vignette(package = "progressr") for examples.

(*) The legacy doSNOW package uses a special foreach() argument .options.doSNOW$progress that can be used to make a progress update each time results from a parallel workers is returned. This approach is limited by how chunking works, requires the developer to set that argument, and the code becomes incompatible with foreach adaptors registered by other doNnn packages.

Examples

plan(multisession)  # parallelize futures on the local machine

y <- foreach(x = 1:10, .combine = rbind) %dofuture% {
  y <- sqrt(x)
  data.frame(x = x, y = y, pid = Sys.getpid())
}
print(y)


## Random number generation
y <- foreach(i = 1:3, .combine = rbind, .options.future = list(seed = TRUE)) %dofuture% {
  data.frame(i = i, random = runif(n = 1L)) 
}
print(y)

## Random number generation (alternative specification)
y <- foreach(i = 1:3, .combine = rbind) %dofuture% {
  data.frame(i = i, random = runif(n = 1L)) 
} %seed% TRUE
print(y)

## Random number generation with the foreach() %:% nested operator
y <- foreach(i = 1:3, .combine = rbind) %:%
       foreach(j = 3:5, .combine = rbind, .options.future = list(seed = TRUE)) %dofuture% {
  data.frame(i = i, j = j, random = runif(n = 1L)) 
}
print(y)

## Random number generation with the nested foreach() calls
y <- foreach(i = 1:3, .combine = rbind, .options.future = list(seed = TRUE)) %dofuture% {
  foreach(j = 3:5, .combine = rbind, .options.future = list(seed = TRUE)) %dofuture% {
    data.frame(i = i, j = j, random = runif(n = 1L)) 
  }
}
print(y)

Description

The doFuture package provides mechanisms for using the foreach package together with the future package such that foreach() parallelizes via any future backend.

Usage

There are two alternative ways to use this package:

1. y <- foreach(...) %dofuture% { ... }

2. y <- foreach(...) %dopar% { ... } with registerDoFuture()

The first alternative (recommended), which uses %dofuture%, avoids having to use registerDoFuture(). The %dofuture% operator provides a more consistent behavior than ⁠%dopar%⁠, e.g. there is a unique set of foreach arguments instead of one per possible adapter. Identification of globals, random number generation (RNG), and error handling is handled by the future ecosystem, just like with other map-reduce solutions such as future.apply and furrr. An example is:

library(doFuture)
plan(multisession)

y <- foreach(x = 1:4, y = 1:10) %dofuture% {
  z <- x + y
  slow_sqrt(z)
}

This alternative is the recommended way to let foreach() parallelize via the future framework if you start out from scratch.

See %dofuture% for more details and examples on this approach.

The second alternative is based on the traditional foreach approach where one registers a foreach adapter to be used by ⁠%dopar%⁠. A popular adapter is doParallel::registerDoParallel(), which parallelizes on the local machine using the parallel package. This package provides registerDoFuture(), which parallelizes using the future package, meaning any future-compliant parallel backend can be used. An example is:

library(doFuture)
registerDoFuture()
plan(multisession)

y <- foreach(x = 1:4, y = 1:10) %dopar% {
  z <- x + y
  slow_sqrt(z)
}

This alternative is useful if you already have a lot of R code that uses ⁠%dopar%⁠ and you just want to switch to using the future framework for parallelization. Using registerDoFuture() is also useful when you wish to use the future framework with packages and functions that uses foreach() and ⁠%dopar%⁠ internally, e.g. caret, plyr, NMF, and glmnet. It can also be used to configure the Bioconductor BiocParallel package, and any package that rely on it, to parallelize via the future framework.

See registerDoFuture() for more details and examples on this approach.

Description

The registerDoFuture() function makes the %dopar% operator of the foreach package to process foreach iterations via any of the future backends supported by the future package, which includes various parallel and distributed backends. In other words, if a computational backend is supported via the Future API, it'll be automatically available for all functions and packages making using the foreach framework. Neither the developer nor the end user has to change any code.

Usage

registerDoFuture()

Value

registerDoFuture() returns, invisibly, the previously registered foreach ⁠%dopar%⁠ backend.

Parallel backends

To use futures with the foreach package and its %dopar% operator, use doFuture::registerDoFuture() to register doFuture to be used as a ⁠%dopar%⁠ adapter. After this, ⁠%dopar%⁠ will parallelize with whatever future backend is set by future::plan().

The built-in future backends are always available, e.g. sequential (sequential processing), multicore (forked processes), multisession (background R sessions), and cluster (background R sessions on local and remote machines). For example, plan(multisession) will make ⁠%dopar%⁠ parallelize via R processes running in the background on the local machine, and plan(cluster, workers = c("n1", "n2", "n2", "n3")) will parallelize via R processes running on external machines.

Additional backends are provided by other future-compliant packages. For example, the future.batchtools package provides support for high-performance compute (HPC) cluster schedulers such as SGE, Slurm, and TORQUE / PBS. As an illustration, plan(batchtools_slurm) will parallelize by submitting the foreach iterations as tasks to the Slurm scheduler, which in turn will distribute the tasks to one or more compute nodes.

Global variables and packages

Unless running locally in the global environment (= at the R prompt), the foreach package requires you do specify what global variables and packages need to be available and attached in order for the "foreach" expression to be evaluated properly. It is not uncommon to get errors on one or missing variables when moving from running a res <- foreach() %dopar% { ... } statement on the local machine to, say, another machine on the same network. The solution to the problem is to explicitly export those variables by specifying them in the .export argument to foreach::foreach(), e.g. foreach(..., .export = c("mu", "sigma")). Likewise, if the expression needs specific packages to be attached, they can be listed in argument .packages of foreach().

When using registerDoFuture(), the above becomes less critical, because by default the Future API identifies all globals and all packages automatically (via static code inspection). This is done exactly the same way regardless of future backend. This automatic identification of globals and packages is illustrated by the below example, which does not specify .export = c("my_stat"). This works because the future framework detects that function my_stat() is needed and makes sure it is exported. If you would use, say, cl <- parallel::makeCluster(2) and doParallel::registerDoParallel(cl), you would get a run-time error on Error in { : task 1 failed - \"could not find function "my_stat" ....

Having said this, note that, in order for your "foreach" code to work everywhere and with other types of foreach adapters as well, you may want to make sure that you always specify arguments .export and .packages.

Load balancing ("chunking")

Whether load balancing ("chunking") should take place or not can be controlled by specifying either argument ⁠.options.future = list(scheduling = <ratio>)⁠ or ⁠.options.future = list(chunk.size = <count>)⁠ to foreach().

The value chunk.size specifies the average number of elements processed per future ("chunks"). If +Inf, then all elements are processed in a single future (one worker). If NULL, then argument future.scheduling is used.

The value scheduling specifies the average number of futures ("chunks") that each worker processes. If 0.0, then a single future is used to process all iterations; none of the other workers are not used. If 1.0 or TRUE, then one future per worker is used. If 2.0, then each worker will process two futures (if there are enough iterations). If +Inf or FALSE, then one future per iteration is used. The default value is scheduling = 1.0.

The name of foreach() argument .options.future follows the naming conventions of the doMC, doSNOW, and doParallel packages, This argument should not be mistaken for the R options of the future package.

For backward-compatibility reasons with existing foreach code, one may also use arguments ⁠.options.multicore = list(preschedule = <logical>)⁠ and ⁠.options.snow = list(preschedule = <logical>)⁠ when using doFuture. .options.multicore = list(preschedule = TRUE) is equivalent to .options.future = list(scheduling = 1.0) and .options.multicore = list(preschedule = FALSE) is equivalent to .options.future = list(scheduling = +Inf). and analogously for .options.snow. Argument .options.future takes precedence over argument .option.multicore which takes precedence over argument .option.snow, when it comes to chunking.

Random Number Generation (RNG)

The doFuture adapter registered by registerDoFuture() does not itself provide a framework for generating proper random numbers in parallel. This is a deliberate design choice based on how the foreach ecosystem is set up and to align it with other foreach adapters, e.g. doParallel. To generate statistically sound parallel RNG, it is recommended to use the doRNG package, where the %dorng% operator is used in place of %dopar%. For example,

y <- foreach(i = 1:3) %dorng% {
  rnorm(1)
}

This works because doRNG is designed to work with any type of foreach ⁠%dopar%⁠ adapter including the one provided by doFuture.

If you forget to use ⁠%dorng%⁠ instead of ⁠%dopar%⁠ when the foreach iteration generates random numbers, doFuture will detect the mistake and produce an informative warning.

For package developers

Please refrain from modifying the foreach backend inside your package or functions, i.e. do not call any registerNnn() in your code. Instead, leave the control on what backend to use to the end user. This idea is part of the core philosophy of the foreach framework.

However, if you think it necessary to register the doFuture backend in a function, please make sure to undo your changes when exiting the function. This can be done using:

  oldDoPar <- registerDoFuture()
  on.exit(with(oldDoPar, foreach::setDoPar(fun=fun, data=data, info=info)), add = TRUE)
  [...]

This is important, because the end-user might have already registered a foreach backend elsewhere for other purposes and will most likely not known that calling your function will break their setup. Remember, your package and its functions might be used in a greater context where multiple packages and functions are involved and those might also rely on the foreach framework, so it is important to avoid stepping on others' toes.

Reporting on progress

How to report on progress is a frequently asked question, especially in long-running tasks and parallel processing. The foreach framework does not have a built-in mechanism for progress reporting(*).

When using doFuture, and the Futureverse in general, for processing, the progressr package can be used to signal progress updates in a near-live fashion. There is special argument related to foreach() or doFuture to achieve this. Instead, one calls a a, so called, "progressor" function within each iteration. See the progressr package and its vignette(package = "progressr") for examples.

(*) The legacy doSNOW package uses a special foreach() argument .options.doSNOW$progress that can be used to make a progress update each time results from a parallel workers is returned. This approach is limited by how chunking works, requires the developer to set that argument, and the code becomes incompatible with foreach adaptors registered by other doNnn packages.

Examples

library(iterators)  # iter()
registerDoFuture()  # (a) tell %dopar% to use the future framework
plan(multisession)  # (b) parallelize futures on the local machine


## Example 1
A <- matrix(rnorm(100^2), nrow = 100)
B <- t(A)

y1 <- apply(B, MARGIN = 2L, FUN = function(b) {
  A %*% b
})

y2 <- foreach(b = iter(B, by = "col"), .combine = cbind) %dopar% {
  A %*% b
}
stopifnot(all.equal(y2, y1))



## Example 2 - Chunking (4 elements per future [= worker])
y3 <- foreach(b = iter(B, by = "col"), .combine = cbind,
              .options.future = list(chunk.size = 10)) %dopar% {
  A %*% b
}
stopifnot(all.equal(y3, y1))


## Example 3 - Simulation with parallel RNG
library(doRNG)

my_stat <- function(x) {
  median(x)
}

my_experiment <- function(n, mu = 0.0, sigma = 1.0) {
  ## Important: use %dorng% whenever random numbers
  ##            are involved in parallel evaluation
  foreach(i = 1:n) %dorng% {
    x <- rnorm(i, mean = mu, sd = sigma)
    list(mu = mean(x), sigma = sd(x), own = my_stat(x))
  }
}

## Reproducible results when using the same RNG seed
set.seed(0xBEEF)
y1 <- my_experiment(n = 3)

set.seed(0xBEEF)
y2 <- my_experiment(n = 3)

stopifnot(identical(y2, y1))

## But only then
y3 <- my_experiment(n = 3)
str(y3)
stopifnot(!identical(y3, y1))

Description

Evaluates a foreach ⁠%dopar%⁠ expression with the doRNG adapter

Usage

withDoRNG(expr, substitute = TRUE, envir = parent.frame())

Arguments

Details

This function is useful when there is a foreach ⁠%dopar%⁠ expression that uses the random-number generator (RNG). Such code should ideally use ⁠%doRNG%⁠ of the doRNG package instead of ⁠%dopar%⁠. Alternatively, and second best, is if the code would temporarily register the doRNG foreach adapter. If neither is done, then there is a risk that the random numbers are not statistically sound, e.g. they might be correlated. For what it is worth, the doFuture adapter, which is set by registerDoFuture(), detects when doRNG is forgotten, and produced an informative warning reminding us to use doRNG.

If you do not have control over the foreach code, you can use withDoRNG() to evaluate the foreach statement with doRNG::registerDoRNG() temporarily set.

Value

The value of expr.

Examples

Consider a function:

my_fcn <- function(n) {
  y <- foreach(i = seq_len(n)) %dopar% {
    stats::runif(n = 1L)
  }
  mean(unlist(y))
}

This function generates random numbers, but without involving doRNG, which risks generating poor randomness. If we call it as-is, with the doFuture adapter, we will get a warning about the problem:

> my_fcn(10)
[1] 0.5846141
Warning message:
UNRELIABLE VALUE: One of the foreach() iterations ('doFuture-1')
unexpectedly generated random numbers without declaring so. There is a
risk that those random numbers are not statistically sound and the overall
results might be invalid. To fix this, use '%dorng%' from the 'doRNG'
package instead of '%dopar%'. This ensures that proper, parallel-safe
random numbers are produced via the L'Ecuyer-CMRG method. To disable this
check, set option 'doFuture.rng.onMisuse' to "ignore".
>

To fix this, we use withDoRNG() as:

> withDoRNG(my_fcn(10))
[1] 0.535326