Resolve workers_within for a distribution spec — resolve

Determines how many parallel workers or callr processes will run inside each job, returning both the resolved count and a human-readable explanation of how it was determined. This makes the auto-detection logic fully inspectable.

Usage

resolve_workers(dist, n_groups = NULL)

Arguments

dist: A parade_dist object (from dist_local(), dist_slurm(), etc.), or NULL.
n_groups: Integer; number of groups assigned to the job/chunk. Used to cap callr workers (the pool can never exceed the number of groups it manages).

Value

A list with components:

workers: Integer — resolved worker count.
source: Short tag: "explicit", "auto", "callr_heuristic", "sequential", "explicit_capped".
detail: Human-readable explanation of how the value was derived.

Examples

# Explicit
resolve_workers(dist_local(within = "callr", workers_within = 10L))
#> $workers
#> [1] 10
#> 
#> $source
#> [1] "explicit"
#> 
#> $detail
#> [1] "workers_within=10 (user-specified)"
#> 

# Auto-detected (will probe your current environment)
resolve_workers(dist_local(within = "multisession"))
#> $workers
#> [1] 4
#> 
#> $source
#> [1] "auto"
#> 
#> $detail
#> [1] "detected 4 workers (parallelly::availableCores())"
#> 

# Shows capping when groups < workers
resolve_workers(
  dist_slurm(by = "subject", within = "callr", workers_within = 20L),
  n_groups = 8L
)
#> $workers
#> [1] 8
#> 
#> $source
#> [1] "explicit_capped"
#> 
#> $detail
#> [1] "workers_within=20 but only 8 groups in chunk; pool capped at 8"
#>