Skip to contents

Changelog

Source: NEWS.md

bio 0.3.2

RStudio settings and dictionaries

  • Preference-file resets are now transactional: failed preset reads or writes restore the original file exactly, or remove partial output when no original file existed. Errors encountered while applying preferences are no longer hidden. The rollback is applied only outside a live RStudio session; a running IDE owns rstudio-prefs.json, so the file is neither deleted nor restored while it is running.

  • The theme change and the ~/R/main directory are only applied when the preference reset actually succeeded.

  • A preference key the running RStudio version does not accept no longer aborts the whole reset. Rejected keys are collected and reported in one warning, and the remaining keys are still applied. Unreadable preset JSON still aborts and rolls back.

  • secure = FALSE now warns that the dictionary archive is fetched over an unencrypted connection and extracted into your RStudio configuration directory.

  • Dictionary archives containing entries that would escape the target directory (absolute paths, drive letters, or .. segments) are refused before extraction.

  • A failed dictionary download now says why the archive was rejected — no data, not a readable zip, unsafe entries, or a missing locale — instead of always reporting an incomplete download. Retries also back off between attempts.

  • The check for the newest available R version discovers the CRAN source directories instead of assuming R 4, and its filename pattern now requires literal dots.

  • rstudio_install_spellcheck_dictionaries() now works outside RStudio by downloading and validating Posit’s dictionary archive, retrying interrupted transfers, and falling back to the system curl command when available. It reports progress through UI messages and invisibly returns its success status. rstudio_download_spellcheck_dictionaries() remains a compatibility alias.

  • Dictionary installation no longer reports success when extraction fails. utils::unzip() signals extraction problems as warnings rather than errors, so the result is now confirmed against the dictionary files that actually reached the target directory.

  • RStudio Desktop detection no longer mistakes the local runtime state directory for an installation directory.

  • rstudio_configure_defaults() no longer creates or deletes ~/Desktop/BS-pratybos/. Deleting a student’s course folder without confirmation was unsafe, so the step was dropped; create the folder manually if you need it. ~/R/main is still created by rstudio_reset_user_settings() for the bio-* presets.

Reliability and maintenance

  • Added three Quarto vignettes (bio, checking-your-setup, rstudio-settings), each marked lifecycle: experimental. Their code chunks are not evaluated, and the badge is vendored so the vignettes render without network access.

  • Expanded the package description and replaced the broad backup.tools namespace import with package-qualified calls and the single narrow import required for the open_backup_dir() re-export.

  • Removed obsolete global-variable declarations, modernized interactive examples, and clarified the generated ..Rcheck/ ignore rule.

  • Online R, RStudio, and Quarto version checks now treat connectivity, transport, parsing, and unexpected-response failures as unavailable version information instead of aborting program checks.

  • GitHub Actions now tests installed-package behavior across current, devel, and older R releases. Package publication and generated-documentation jobs are serialized, deployment failures remain visible, and documentation jobs follow the current R release rather than a hardcoded patch version. Generated documentation is committed back to same-repository pull-request branches; read-only fork pull requests receive a patch artifact.

  • Regression coverage was expanded for preference rollback, confirmation and reset summaries, malformed version responses, dictionary installation, bundled JSON assets, and deterministic internal helpers.

  • The test suite no longer reaches the real network: the connectivity probe is mocked everywhere an online version check is exercised, so results no longer depend on runner connectivity.

  • The lint workflow now fails when lintr reports findings (LINTR_ERROR_ON_LINT), and the package is lint-clean. .lintr documents why object_usage_linter is disabled (glue interpolation through usethis::ui_*()) and allows camelCase names required by external APIs.

bio 0.3.1

rstudio_compare_user_settings() improvements

  • rstudio_compare_user_settings() prints a short, readable summary by default (how many settings match, plus a grouped list of what differs) instead of a raw waldo::compare() dump. Use the new output argument to control verbosity: "concise" (default), "minimal" (counts only), or "verbose" (full waldo::compare() output for debugging).

  • New source argument ("auto", "live", "file") controls how “current” settings are read: live via rstudioapi (requires a running RStudio session), from the saved rstudio-prefs.json file on disk, or automatically picking whichever is available.

  • rstudio_compare_user_settings() now works without a running RStudio session, comparing against the saved rstudio-prefs.json file instead of failing. Settings left at RStudio’s built-in default (and therefore absent from that file) are backfilled from the locally installed RStudio’s preference schema, so they’re no longer misreported as “missing” — including nested settings like panes$hiddenTabSet.

Headless RStudio support

  • rstudio_reset_user_settings() and rstudio_set_preferences() now work when called outside RStudio (e.g. via Rscript), merging preference JSON files directly instead of requiring a live rstudioapi session.

  • rstudio_reset_gmc() was split into rstudio_configure_defaults() (headless-capable: preferences, keybindings, snippets, directories, dictionaries) and rstudio_reset_session_state() (requires a live RStudio session: tabs, workspace, layout, theme, console, history). rstudio_reset_gmc() remains as a thin wrapper calling both, for backward compatibility.

  • Added internal helpers for detecting an installed (not necessarily running) RStudio Desktop and its version, used to support the above.

  • New get_rstudio_install_scope() reports whether the local RStudio Desktop install is system-wide (“all users”) or per-user (“just me”). find_rstudio_install_dir() now looks for both, checking the per-user registry hive on Windows that it previously missed.

Other changes

  • Documentation and AI-assistant context files updated to match current behavior.

  • Fixed rstudio_delete_spellcheck_dictionaries(ask = FALSE), which could previously fail before removing the selected dictionary directory.

  • Fixed open_project() when called with a supplied proj_list and no proj_list_path.

  • Fixed get_path_rstudio_config_dir() when XDG_CONFIG_HOME is set by appending the rstudio sub-directory per the XDG Base Directory specification.

  • Added isolated regression coverage for dictionary deletion and canceled RStudio user-settings resets.

  • Added mocked regression coverage for RStudio restart/reload commands and spellcheck dictionary installation.

  • Expanded automated coverage across exported package, path, project, file opening, RStudio integration, and offline program-information helpers.

  • Fixed clear_r_history() to use RStudio’s history commands in a live RStudio session, including on Windows.

  • Clarified that the bundled rstudio-default preset was compiled from the most recent official RStudio settings documentation available in 2022; setting comparisons use the locally installed RStudio schema when available.

  • Added regression coverage for open_project() availability filtering and duplicate project names outside an interactive RStudio session.

  • Updated RStudio Desktop support links to Posit’s current support site and corrected current Linux/macOS internal-state paths to ~/.local/share/rstudio.

  • Updated remaining RStudio documentation links to docs.posit.co and documented the unverified Windows/RStudio command-history reset behavior.

  • Fixed RStudio install-scope classification tests on non-Windows runners and made per-user path detection separator- and path-boundary-safe.

  • Removed retired, commented-out implementation experiments from the package source.