Skip to content

writeAlizer 1.7.0

Choose a tag to compare

View all tags
@shmercer shmercer released this 21 Oct 19:52
· 40 commits to master since this release
v1.7.0
fdb2656

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.
Assets 3
Loading