fapply() and every mapper (fmap(),
fmapn(), fmapr(), fmapc(),
fmapg(), fwalk(), fcv(),
frepeat()) gain .on_error: "stop"
(default, unchanged behaviour), "pass" (continue and store
a functionals_error object in the failed position, then
emit a one-line summary warning), or "fill" (substitute
.fill). The policy behaves identically in sequential, fork,
and socket-cluster execution..seed argument for reproducible iteration. Each
task runs with its own L’Ecuyer-CMRG random-number stream, so results do
not depend on ncores or task-completion order. The caller’s
global RNG state is saved and restored.
frepeat(..., .seed =) makes parallel Monte Carlo runs
reproducible.system.file("benchmarks/bench-purrr.R", package = "functionals").floop() is deprecated. It never did anything beyond
fmap() (collect results) and fwalk() (side
effects only); it now warns and forwards to them, and will be removed in
a future release.ncores handling is unified through a single internal
normaliser: NULL, 1, and 1L all
mean sequential; invalid values warn and fall back to sequential,
consistently across every function.fmapn() now forwards ... to
.f instead of leaking the extra arguments into an internal
helper (previously an error whenever ... was
non-empty).Removed the stray LazyData: true field (the package
ships no data/ directory, so it triggered an “Omitted
‘LazyData’” note at build time).
Wrapped long example lines in floop(),
fmapc(), fmapg(), fmapn(), and
frepeat() that were overflowing the PDF manual’s page width
and triggering a LaTeX WARNING under
R CMD check --as-cran.
Simplified DESCRIPTION’s Description:
field: dropped a stray cross-reference to another package by name and a
GitHub link (URL/ BugReports already cover that).
Standardized Authors@R family-name casing and split
URL/BugReports so URL holds only
the CRAN package page and BugReports only the GitHub issues
page, matching the pattern used across all packages.
Fixed fapply() and fmap() progress
reporting so sequential, multicore, and cluster-backed execution advance
on completed tasks rather than internal chunk boundaries.
Improved progress-bar output stability when user code emits messages during tracked execution.
Added progress-system regression tests covering sequential timing, exact parallel counts, and error propagation in cluster and multicore execution.
Updated package documentation to describe completion-driven progress semantics.