Create a parade flow for declarative data processing

A flow defines a computational pipeline with typed stages that operate on a parameter grid. Each stage can depend on previous stages and produce typed outputs with optional error handling policies.

Usage

flow(
  grid,
  seed_col = NULL,
  error = c("propagate", "keep", "omit", "stop"),
  retries = 0L,
  retry_backoff = c("none", "fixed", "linear", "exponential"),
  retry_base = 0,
  retry_on = NULL,
  cancel = c("deps", "all", "none"),
  resources = NULL,
  cpus = NULL,
  memory = NULL,
  time = NULL
)

Arguments

grid: A data frame or tibble containing parameter combinations
seed_col: Optional column name for reproducible random seeds
error: Error handling policy: "propagate" (default), "keep", "omit", or "stop"
retries: Flow-level retry attempts per stage (default 0).
retry_backoff: Backoff strategy for retries: "none", "fixed", "linear", "exponential".
retry_base: Base delay in seconds for retry backoff.
retry_on: Optional retry classifier. NULL retries all errors; a character vector retries matching condition classes; a function receives the condition and should return TRUE/FALSE.
cancel: Cancellation propagation mode used when error = "propagate": "deps" (default), "all", or "none".
resources: Optional flow-level resource defaults as a named list with keys cpus, memory, time (aliases: cpus_per_task, ncpus, mem).
cpus: Optional flow-level default CPUs per task.
memory: Optional flow-level default memory (e.g. "8G").
time: Optional flow-level default walltime (e.g. "2:00:00").

Value

A parade_flow object containing the grid, stages, and options

Examples

# Create a simple flow
grid <- data.frame(x = 1:3, y = letters[1:3])
fl <- flow(grid)
print(fl)
#> <parade_flow>
#>   Grid     : 3 rows × 2 cols [x, y]
#>   Stages   : 0 []
#>   Error    : propagate

# Flow with seed column for reproducibility
fl_seed <- flow(grid, seed_col = "x")