[psychelzh]

Summary

• replace the legacy threshold/k-fold API and the old fit()/fit_resamples() split with a single staged workflow built around spec() + cpm()
• standardize on one fold-based cpm result object and simplify summary/tidy output to correlation-focused reporting
• simplify the runtime/file layout, remove single-fit leftovers, and align docs, tests, pkgdown, and touchstone with the current workflow

API changes

• spec() and cpm() are now the only public workflow entry points
• remove cpm_spec(), fit(), fit_resamples(), cpm_resamples(), and resample_metrics()
• stage helpers are now cpm_selection_cor(), cpm_construction_strength(), and cpm_model_lm()
• cpm() runtime settings now center on spec, resamples, covariates, return_edges, and na_action
• summary() / tidy() now report correlation-based CPM results only

Internal cleanup

• consolidate the main runtime into R/cpm-spec.R, R/cpm.R, R/fit.R, and R/folds.R while keeping selection/construction as explicit CPM stages
• remove the old single-fit path and related helper layers
• realign tests/testthat to the current module boundaries
• update touchstone benchmarks so main and this branch can be compared under the old and new APIs

Validation

• air format .
• devtools::document()
• devtools::build_readme()
• devtools::test() # 312 passing
• pkgdown::build_site(preview = FALSE, install = TRUE)
• covr::package_coverage() # 100%
• devtools::check() # 0 errors / 0 warnings / 0 notes

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (5d7ad83) to head (6a568f3).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##              main       #56    +/-   ##
==========================================
  Coverage   100.00%   100.00%            
==========================================
  Files            4         6     +2     
  Lines          346      1032   +686     
==========================================
+ Hits           346      1032   +686     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[chatgpt-codex-connector]

💡 Codex Review

cpmr/R/cpm.R

Line 214 in 6a4913a

tibble::as_tibble(apply(sum_x$edges, 2, list))

Guard edge tidying when edges are not stored

tidy.cpm(component = "edges") unconditionally calls apply(sum_x$edges, 2, list), but sum_x$edges is NULL for the default cpm() workflow (return_edges = "none"). In that common path this raises a hard error (dim(X) must have a positive length) instead of returning a usable result or a clear user-facing message, so users cannot tidy edge output unless they know to override edge storage up front.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[psychelzh]

/benchmark

[github-actions]

Touchstone benchmark results for codex/cpm-unified-workflow versus main (with a 95% confidence interval in relative change):

• 🚀workflow_default: 272ms -> 235ms [-15.2%, -12.53%]
• 🚀workflow_standardize_edges: 499ms -> 392ms [-22.51%, -20.63%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[psychelzh]

/benchmark

[github-actions]

Touchstone benchmark results for codex/cpm-unified-workflow versus main (with a 95% confidence interval in relative change):

• 🚀workflow_default: 246ms -> 191ms [-24.62%, -20.72%]
• 🚀workflow_standardize_edges: 513ms -> 343ms [-34.44%, -32.06%]
  Further explanation regarding interpretation and methodology can be found in the documentation.