Skip to contents

Changelog

Source: NEWS.md

Version 2026.9.24

Bug fixes

  • run_all() with foreach returned a gc() matrix for every analysis. It now returns what each action function returns.
  • A fn_name now resolves first in the local environment where it resolved when you added it, then in the global environment and on the search path, and last in plnr. A plnr function no longer wins over your function of the same name. The plan stores only that local environment. get_anything() gains envir and mode.
  • The vignettes now knit with rmarkdown::render(envir = new.env()). They failed with object 'fn_fig_1' not found.
  • progress is now in Imports. The default verbose path of run_all() called it without declaring it.
  • .plnr.options = list(chunk_size = n) now reaches foreach. It no longer reaches the action function.
  • add_argset() and add_analysis() no longer copy the whole list of analyses on every call.
  • create_rmarkdown() now ends every file that it writes with a newline.
  • run_all() with foreach on a PSOCK cluster, such as one from doParallel, failed with object 'private' not found. The foreach call now exports what the workers need.

Behaviour changes

  • add_argset_from_df() and add_analysis_from_df() with zero rows add nothing. They failed, and left an analysis named NA.
  • get_argsets_as_dt() reads the element argset exactly. An element named argsets no longer counts as the argset.
  • With use_foreach = NULL and one registered worker, run_all() no longer loads the progressr namespace.
  • Errors that plnr raises itself, such as “Both fn and fn_name are NULL”, carry no call. The messages are unchanged.
  • try_again() evaluates verbose only when it reports. try_again(1, verbose = stop("forced")) now returns TRUE. It failed.
  • Without progressr, run_all() with foreach and verbose = TRUE runs without a progress bar. It failed. run_all_progress() without progressr now stops with a message that names the package.

Checks

  • The code passes the static-checks lint gate again. It had failed since 2026-08-26, so pkgdown had not deployed since 2026.8.21.
  • R/plan.R is exempt from two lint rules, cyclocomp_linter and vector_logic_linter, because their fixes change behaviour.

Documentation

  • The help pages, both vignettes, README.md, index.md and the DESCRIPTION text follow ASD-STE100.
  • No page claims hash-based caching. get_data() computes digests but plnr never reads them.
  • The help pages now say that ... goes to the action function, and give the true return values.

Version 2026.9.23

CRAN release: 2026-09-23

  • This CRAN release carries the Plan$add_analysis_from_list() fix from 2026.8.3 (#1). CRAN version 2025.11.22 does not have it.
  • DESCRIPTION now declares R (>= 4.1.0), because the code uses the base pipe |>.

Version 2026.8.21

  • The package drops magrittr. Every %>% is now the base pipe |>, and magrittr is gone from DESCRIPTION.
  • The rewrite is a relocation, not an edit. Each %>% call was transformed the way R’s parser transforms |>, and the resulting tree was required to match the tree parsed from the rewritten file. A file whose trees disagreed was left untouched and converted by hand instead.

Version 2026.8.6

Licensing

  • DESCRIPTION Authors@R now declares Richard Aubrey White as the copyright holder, with role = "cph". It declared none at all, and neither did any other package in the fleet. Nothing in R CMD check reports that.
  • The copyright year is now 2026. It read 2025.
  • CLAUDE.md now carries a Licensing section, so the year gets checked rather than silently ageing.

Corrections

  • try_again() was titled “Retry code execution with exponential backoff”. It does not implement exponential backoff. The delay is stats::runif(1, delay_seconds_min, delay_seconds_max), drawn afresh before every retry from fixed bounds that the loop never reassigns, so it does not grow with the attempt number. The title now names the real behaviour, and the description says what the delay is and what it is not. No code changed, and the arguments and their defaults are untouched.

  • Prose only. This release rewrites the roxygen documentation, both vignettes, README.md, index.md and NEWS.md to ASD-STE100 (Simplified Technical English). The R sources are semantically identical to the previous release.

  • No claim changed. The sweep found documented claims that the code does not support. It left every one of those claims in place, and reported it.

  • The rewrite splits long sentences, prefers the active voice, and uses one term for each concept. It also uses the RFC-2119 keywords MUST and MAY in the test_action_fn() @seealso, which states the action-function contract.

Version 2026.8.3

  • This release is mostly documentation and roxygen comments. It also carries two bug fixes that do change behavior: create_rmarkdown() and Plan$add_analysis_from_list(). The entries below describe both. Apart from those two fixes, the R sources are semantically identical to the previous release.
  • @seealso added to all ten exported functions. Eight point at the vignette that covers them. No vignette covers set_opts() or try_again(), and the text of those two says so.
  • @family example and test functions added to example_action_fn(), test_action_fn() and example_data_fn_nor_covid19_cases_by_time_location().
  • Runnable examples added to example_data_fn_nor_covid19_cases_by_time_location() and test_action_fn(), which previously had none.
  • try_again() examples taken out of \dontrun{}. They now run, and show a retry that fails once and succeeds on the second attempt.
  • create_rmarkdown() examples moved from \dontrun{} to \donttest{}, so R CMD check runs them.
  • Fixed create_rmarkdown(), which evaluated {lubridate::today()} while it generated run.R, instead of writing the brace out literally. The brace was not escaped, so glue resolved it at generation time. That had two consequences. Every generated project carried a frozen date, not one that evaluates when the user runs it. create_rmarkdown() also silently needed lubridate, which plnr does not depend on. Both consequences are gone.
  • index.md and Rplots.pdf added to .Rbuildignore, so R CMD build no longer ships the pkgdown home page source in the tarball.
  • Fixed Plan$add_analysis_from_list(). It decided whether to apply the method-level fn_name by reading names(df). df is not a parameter of that method, and the method does not define it. The lookup therefore escaped to whatever df the search path offered, normally stats::df, the F-distribution density. names(stats::df) is NULL, so the guard always passed and the mistake stayed invisible. The method now reads names(argset). That is what the line above it builds, and what the sibling add_analysis_from_df() guards on. The fix has two consequences. An argset that carries its own fn_name keeps it, instead of losing it to a silent overwrite, which matches add_analysis_from_df(). The method also no longer fails when stats is not attached, which is how R CMD check loads a package in its minimal-namespace tests. tests/testthat/test-add-analysis-from-list.R covers the fix.

Version 2025.11.22

CRAN release: 2025-11-22

Version 2025.3.19

  • Improvements to roxygen2 documentation across all R files:
    • Restructured the Plan class documentation and made the explanations clearer
    • Added detailed parameter descriptions and examples to the method documentation
    • Added documentation for the internal functions
    • Added documentation for the utility functions is_run_directly(), get_anything() and expand_list()
    • Added detailed examples and usage patterns for all functions
    • Made all the documentation clearer and more consistent

Version 2024.1.18

  • Created the plan$run_all_parallel() public method.

Version 2022.6.8

CRAN release: 2022-06-09

  • Created the plan$set_use_foreach() public method.

Version 2022.6.7

  • CRAN submission.
  • Additional documentation.
  • Moved some plan public fields into private.

Version 2022.5.27

  • Added a new vignette that explains how to add analyses to a plan.

Version 2022.4.6

  • Added hash functions to get_data().

Version 2021.6.9

  • Added easy_split().

Version 2020.5.11

  • fn_name can now take package::function_name arguments.

Version 2020.5.4

  • Created the is_run_directly() function. It shows the user whether their code runs directly, or from within a function.

Version 2020.4.3

  • Created the set_opts() function. It sets force_verbose package wide.

Version 2020.2.20

Version 2020.1.28

  • Added the parallel_possible variable for a new Plan.