writeAlizer 1.7.2

new features

• Added keep_stem_before_txt() function to process Coh-Metrix and GAMET filenames that appear as paths
• integrated keep_stem_before_txt() into import_coh() import_gamet() and import_merge_gamet_rb() functions

improved documentation

• Added new vignette that details scoring model development
• Added a pkgdown() site to host documentation on GitHub

writeAlizer 1.7.1

writeAlizer 1.7.1 (2025-10-22)

CRAN compliance and offline example guard

• Fixed predict_quality() example to run deterministically when options(writeAlizer.offline = TRUE).
• Example now seeds a local 'example' model with wa_seed_example_models("example", dir = tempdir()) and restores prior options after cleanup.
• Offline runs now fully skip remote file downloads, avoiding spurious errors during R CMD check --as-cran --run-donttest.
• No user-facing API changes.

writeAlizer 1.7.0

writeAlizer 1.7.0 (2025-10-21)

Error handling & classes

• Introduced structured errors via rlang::abort():
  • writeAlizer_registry_missing – missing/invalid artifacts.csv.
  • writeAlizer_unknown_model – unknown/unsupported model key (raised by preprocess() / predict_quality()).
  • writeAlizer_download_failed – offline/bad URL/checksum fetch failures.
  • writeAlizer_internal_mismatch – expected submodel count ≠ preprocessing splits.
  • .wa_ensure_file - added classed errors and download retries

Loaders & caching

• .wa_ensure_file():
  • Honors options(writeAlizer.mock_dir=) (preserves subdirs).
  • Supports file:// URLs; clearer offline guidance.
  • Warns on cached checksum mismatch; hard-errors if post-download checksum mismatches.
• .wa_parts_for() validates inputs and uses canonicalized model keys.
• .wa_load_model_rdas() / .wa_load_fits_list():
  • Prefer mock_dir artifacts; load via temp env; pick sensible object (fit, model, etc.).
  • Legacy → v2 name mapping; special-case example.
• .wa_require_pkgs_for_fits() collects deps from model classes and train$modelInfo$library; errors with install hint.
• .wa_validate_import() - checks ID column for duplicates and character type

model_deps()

• Reports required vs missing from Suggests + override options(writeAlizer.required_pkgs=).
• Preserves version qualifiers in required; strips qualifiers for missing.
• Emits copy-paste install.packages() command.
• Tests updated to capture output/messages cleanly (no stray console prints).

Predict API & preprocessing

• Better validation for swapped args and NA/invalid model with actionable messages.
• predict_quality() surfaces unknown model keys (includes canonicalized key).
• Preprocessing preserves row counts/ID; supports v2 varlists and consolidated-split behavior.

Test infrastructure

• Added wa_sample_path() to reliably find inst/extdata in dev/install contexts.
• Network helper hardened; integration tests force offline via options(writeAlizer.offline=TRUE) and avoid HEAD probes.
• Introduced .quiet_eval() to silence messages/warnings in targeted assertions.

Integration & smoke tests

• Integration test now mocks the registry (RDS + RDA) for all exercised models:
  • RB: rb_mod1, rb_mod2, rb_mod3narr_v2, rb_mod3exp_v2, rb_mod3per_v2, rb_mod3all_v2
  • Coh: coh_mod1, coh_mod2, coh_mod3narr, coh_mod3exp, coh_mod3per, coh_mod3all
  • GAMET: gamet_cws1
• Writes matching varlists/fits into mock_dir; verifies prediction columns and mean-column rules.
• rb_mod3all smoke test made offline-only with mocked varlists.

RB/COH import utilities tests

• Coverage for:
  • import_rb() path with SEP=, + read.table(text=readLines()).
  • NaN→NA in character columns; guarded numeric-like conversion (accepts integer where appropriate).
  • Name-based keep/drop assertions driven by sample header.
• Skips only when assets truly missing (via wa_sample_path()).

Misc

• Added a package vignette
• switch to MIT license
• Added/updated tests for:
  • Mismatched expected submodels.
  • Bad registry columns.
  • Checksum errors.
  • Legacy → v2 artifact-name mapping.

writeAlizer 1.6.5

writeAlizer 1.6.5

• Addresses CRAN policy for Internet resources:
  • Network access now fails gracefully with informative messages (no low-level download.file errors).
  • Tests that rely on remote artifacts now preflight their exact URLs/paths and skip when unavailable.
  • Examples/tests seed mock artifacts for deterministic offline runs.

writeAlizer 1.6.4

writeAlizer 1.6.4 (2025-09-23)

CRAN policy & documentation improvements

• Replaced writeAlizer:::wa_seed_example_models() with writeAlizer::wa_seed_example_models() in examples and documentation, per CRAN’s guidance to avoid accessing unexported objects.

• Exported wa_seed_example_models() with full documentation:

  • Added @return (returns the created temp path invisibly).
  • Added a small, runnable example that writes only to tempdir() and cleans up; it also restores any prior option value.
  • Documented the temporary side effect of setting options(writeAlizer.mock_dir = <path>).

writeAlizer 1.6.3

writeAlizer 1.6.3 (2025-09-17)

CRAN policy & documentation improvements

• DESCRIPTION: acronyms expanded (e.g., CBM), software/package names single-quoted, and method references added with CRAN auto-link format (DOIs).
• Documentation: @return added to import_coh(), import_rb(), import_gamet(), import_merge_gamet_rb().
• Examples: \\dontrun{} replaced with \\donttest{} where needed; fast examples unwrapped; predict_quality() uses a tiny offline 'example' model seeded in tempdir() with cleanup.
• File I/O: no writes to home by default; examples/tests use tempdir(); caching standardized via tools::R_user_dir("writeAlizer","cache").
• New helpers: wa_cache_dir() and wa_cache_clear(ask=) for cache discovery/cleanup.
• Optional deps: new model_deps() reports (does not install) optional packages and prints an install hint; install_model_deps() deprecated.
• Tests: self-cleaning with withr fixtures; no temp detritus; removed deprecated global teardown.

Bug fixes & maintenance

• Stabilized example/test behavior when offline by mocking artifacts via writeAlizer.mock_dir.
• Suppressed noisy pre-processing variance warnings in integration tests.

writeAlizer 1.6.2

implifies install_model_deps():

• Remove unused model argument.
• Standardize on the option hook: options(writeAlizer.require_pkgs_for_fits = function() ...).
• Collapse discovery to a single utils::packageDescription(..., fields = "Suggests") path.
• dry_run = TRUE now returns tokens visibly; version qualifiers are stripped only for checks/installs.

Tests

• Updated tests to use the options hook and visible dry-run behavior.
• Mocked requireNamespace() and utils::install.packages() to verify missing-package handling.
• All tests pass locally with devtools::test().

Notes

• Version set to 1.6.2 in DESCRIPTION. (Optional Date: field set to YYYY-MM-DD if present.)
• NEWS.md updated with a concise entry.

writeAlizer 1.6.1

writeAlizer 1.6.1

Tests & Coverage

• Added comprehensive tests for artifact registry helpers, cache, and installer.
• Increased overall coverage to ~82% (artifact_registry ~89%, cache ~92%, download 100%).

Artifact registry

• Prefers inst/metadata/artifacts.csv; robust fallbacks and input validation.
• Deterministic part ordering; legacy model key mapping handled.

Downloads

• Suppressed download.file chatter in tests and internal helpers.
• Added checksum verification path in ensure-file.

install_model_deps()

• Dry-run returns canonical Suggests tokens.
• Non-dry-run installs only truly missing packages (strips version tokens).
• Optional helper hook respected via options(writeAlizer.require_pkgs_for_fits=...).
• Added examples and improved docstrings.

Misc

• Quieted noisy test logs; hardened temp-dir handling.

writeAlizer 1.6.0

• Replaced the ReaderBench mod3 models with their v2 counterparts (e.g., rb_mod3all now calls the rb_mod3all_v2 models).
• Optimized model artifact downloads and added SHA-256 checksum verification (auto re-download on mismatch).
• predict_quality() now returns per-submodel scores and the overall mean column pred_<model>_mean (where applicable).
• import_rb() now keeps ReaderBench features by name using the packaged sample_rb.csv header (first 404), dropping 405+. It falls back to the legacy positional selection if the header is unavailable, reducing brittleness to column reordering.
• Added install_model_deps() helper to install Suggests for a given model.
• CI: wired up GitHub Actions R-CMD-check (Windows/macOS/Linux).