Submit a flow for deferred execution

Submits a parade flow for asynchronous execution, either locally using future or on SLURM using batchtools. Returns a handle for monitoring and collecting results.

Usage

submit(
  fl,
  mode = c("index", "results"),
  run_id = NULL,
  registry_dir = NULL,
  index_dir = NULL,
  seed_furrr = TRUE,
  scheduling = 1,
  clean = FALSE,
  label = NULL
)

Arguments

fl: A parade_flow object with distribution settings
mode: Execution mode: "index" (default) or "results"
run_id: Optional run identifier (auto-generated if NULL)
registry_dir: Directory for execution registry
index_dir: Directory for result indices
seed_furrr: Whether to enable deterministic random number generation
scheduling: Furrr scheduling parameter
clean: If TRUE, remove an existing registry directory before creating a new one. Useful when retrying after a failed submission.
label: Optional human-readable submit label. Used to derive the default registry directory name and backend job/task names.

Value

A parade_deferred object for monitoring execution

Examples

# \donttest{
grid <- data.frame(x = 1:4, group = rep(c("A", "B"), 2))
fl <- flow(grid) |>
  stage("calc", function(x) x^2, schema = returns(result = dbl())) |>
  distribute(dist_local(by = "group", within = "sequential"))

deferred <- submit(fl)
#> [parade] submit prune: scanning 2 groups for cached outputs
#> [parade] submit prune complete in 0.0s (0 pruned, 2 pending)
unlink(c(paths_get()$registry, paths_get()$artifacts), recursive = TRUE)
unlink("parade.log")
# }