Map Over Elements With Fingerprinted Checkpoints

A purrr/furrr-style mapping function that caches individual task results to disk under cryptographic fingerprints. Tasks that have already been evaluated into `checkpoint_dir` with a matching fingerprint are loaded from disk and skipped. When `shard = c(i, n)` is supplied, only shard `i` of `n` tasks is evaluated; other tasks are returned from disk if available or as `NULL`.

Usage

checkpoint_map(
  .x,
  .f,
  ...,
  checkpoint_dir = NULL,
  key = NULL,
  fingerprint = NULL,
  shard = NULL,
  checkpoint_mismatch = c("recompute", "overwrite", "error"),
  progress = FALSE
)

Arguments

.x

A list, vector, or data frame of inputs to iterate over.

.f

A function, specified or formula, to apply to each element of `.x`.

...

Additional arguments passed to `.f`.

checkpoint_dir

Optional directory path where per-task `.rds` checkpoint files are stored. If `NULL`, executes normally without disk caching.

key

Optional vector or function producing safe, unique character keys for each element of `.x`. Defaults to names of `.x`, or `item_0001`, `item_0002`, etc.

fingerprint

Optional character vector of pre-computed fingerprints per item, or a function `function(x, ...)` returning a fingerprint string. When `NULL`, fingerprints are derived from `.x[[i]]`, `...`, and package version.

shard

Optional integer vector `c(i, n)` specifying that this execution evaluates shard `i` of `n` items.

checkpoint_mismatch

Action when an existing checkpoint file has a mismatched fingerprint: `"recompute"` (default), `"overwrite"`, or `"error"`.

progress

Logical (default `FALSE`). Display progress indication.

Value

A list of evaluated (or loaded) results with names preserved from `.x`.

See also

[collect_checkpoints()], [clean_checkpoints()], [tune_grid_checkpointed()]

Examples

# \donttest{
tmp <- file.path(tempdir(), "ckpt_map_demo")
res <- checkpoint_map(1:3, function(x) x^2, checkpoint_dir = tmp)
# Second run loads from disk instantaneously
res2 <- checkpoint_map(1:3, function(x) x^2, checkpoint_dir = tmp)
unlink(tmp, recursive = TRUE)
# }