Prepare 0.2.0 for CRAN submission

DESCRIPTION

@@ -7,12 +7,12 @@ Authors@R:
 Description: Filter, wrapper, and embedded feature-selection methods behind a
     consistent set of functions that share one calling convention and one
     return type: univariate threshold filters, correlation and chi-squared
-    filters, information gain, LASSO and elastic net, Bayesian model
-    comparison, Boruta, recursive feature
-    elimination, random forest importance, multivariate adaptive regression
-    splines, support vector machine recursive feature elimination, stepwise
-    selection, and principal component / singular value decomposition
-    helpers. The implemented methods follow
+    filters, information gain, the least absolute shrinkage and selection
+    operator (LASSO) and elastic net, Bayesian model comparison, Boruta,
+    recursive feature elimination, random forest importance, multivariate
+    adaptive regression splines, support vector machine recursive feature
+    elimination, stepwise selection, and principal component / singular
+    value decomposition helpers. The implemented methods follow
     Tibshirani (1996) <doi:10.1111/j.2517-6161.1996.tb02080.x>,
     Zou and Hastie (2005) <doi:10.1111/j.1467-9868.2005.00503.x>,
     Friedman (1991) <doi:10.1214/aos/1176347963>,
@@ -35,7 +35,7 @@ Imports:
     parallel,
     stats,
     utils,
-    withr
+    withr (>= 2.4.0)
 Suggests:
     bigstatsr,
     Boruta,

NEWS.md

@@ -5,6 +5,11 @@ documentation page found and fixed silent wrong answers, hangs, crashes, and
 inaccurate claims. Fixes that change behavior are listed first under each
 review. Where a result changed, the old result was wrong.
 
+* featR now requires withr 2.4.0 or later, the first release that exports
+  `local_seed()`.
+* Examples that run in under two seconds are no longer wrapped in
+  `\donttest{}`, so `R CMD check` runs them.
+
 ## First adversarial review
 
 A full statistical, methodological, and programmatic review. Every finding

R/fs_elastic.R

@@ -660,7 +660,6 @@ elastic_check_categorical_levels <- function(data, predictors) {
 #'     `use_pca = TRUE`).}
 #'   \item{call}{The matched call.}
 #' @examples
-#' \donttest{
 #' if (requireNamespace("caret", quietly = TRUE) &&
 #'     requireNamespace("glmnet", quietly = TRUE) &&
 #'     requireNamespace("Matrix", quietly = TRUE)) {
@@ -672,10 +671,9 @@ elastic_check_categorical_levels <- function(data, predictors) {
 #'   )
 #'   df$y <- 2 * df$x1 - df$x2 + 0.1 * cos(seq_len(60) * 3)
 #'   res <- fs_elastic(df, "y", alpha_seq = c(0.5, 1), seed = 1)
-#'   selected(res)
-#'   res$scores
-#'   res$details$best_alpha
-#' }
+#'   print(selected(res))
+#'   print(res$scores)
+#'   print(res$details$best_alpha)
 #' }
 #' @seealso The [Regularization and embedded importance](https://elkronos.github.io/featR/articles/regularization.html)
 #'   guide, which runs this function on a worked example.

R/fs_lasso.R

@@ -531,7 +531,6 @@ lasso_importance <- function(coefs) {
 #'     and `n_features` (number of design-matrix columns considered).}
 #'   \item{call}{The matched call.}
 #' @examples
-#' \donttest{
 #' if (requireNamespace("glmnet", quietly = TRUE)) {
 #'   n <- 100
 #'   df <- data.frame(
@@ -541,9 +540,8 @@ lasso_importance <- function(coefs) {
 #'   )
 #'   df$y <- 2 * df$x1 - 3 * df$x2 + rnorm(n)
 #'   res <- fs_lasso(df, "y", seed = 123)
-#'   selected(res)
-#'   head(res$scores)
-#' }
+#'   print(selected(res))
+#'   print(head(res$scores))
 #' }
 #' @seealso The [Regularization and embedded importance](https://elkronos.github.io/featR/articles/regularization.html)
 #'   guide, which runs this function on a worked example.

R/fs_mars.R

@@ -733,7 +733,6 @@ mars_evaluate <- function(model, test, target, verbose = FALSE) {
 #' }
 #'
 #' @examples
-#' \donttest{
 #' if (requireNamespace("caret", quietly = TRUE) &&
 #'     requireNamespace("earth", quietly = TRUE)) {
 #'   # x1 and x2 drive y; x3 is noise
@@ -745,10 +744,9 @@ mars_evaluate <- function(model, test, target, verbose = FALSE) {
 #'   df$y <- 2 * df$x1 - df$x2 + rnorm(150, sd = 0.5)
 #'   res <- fs_mars(df, "y", degree = 1, nprune = c(5, 10),
 #'                  number = 3, repeats = 1, seed = 42)
-#'   res$selected
-#'   res$scores
-#'   res$details$metrics
-#' }
+#'   print(res$selected)
+#'   print(res$scores)
+#'   print(res$details$metrics)
 #' }
 #' @seealso The [Regularization and embedded importance](https://elkronos.github.io/featR/articles/regularization.html)
 #'   guide, which runs this function on a worked example.

R/fs_pca.R

@@ -494,7 +494,6 @@ pca_plot <- function(pca_result, label_col) {
 #' res$var_explained
 #' head(res$pca_df)
 #'
-#' \donttest{
 #' if (requireNamespace("ggplot2", quietly = TRUE)) {
 #'   # plot = FALSE (the default) draws nothing, but $plot is built anyway
 #'   res <- fs_pca(iris, label_col = "Species", verbose = TRUE)
@@ -503,7 +502,6 @@ pca_plot <- function(pca_result, label_col) {
 #'   # plot = TRUE draws it as well
 #'   res <- fs_pca(iris, label_col = "Species", plot = TRUE)
 #' }
-#' }
 #' @seealso The [Dimensionality reduction](https://elkronos.github.io/featR/articles/dimensionality-reduction.html)
 #'   guide, which runs this function on a worked example.
 #' @export

R/fs_randomforest.R

@@ -401,7 +401,6 @@ rf_pool_importance_sd <- function(forests) {
 #' }
 #'
 #' @examples
-#' \donttest{
 #' if (requireNamespace("randomForest", quietly = TRUE) &&
 #'     requireNamespace("caret", quietly = TRUE)) {
 #'   res <- fs_randomforest(
@@ -416,7 +415,6 @@ rf_pool_importance_sd <- function(forests) {
 #'   print(res$scores)
 #'   print(res$details$metrics)
 #' }
-#' }
 #' @seealso The [Regularization and embedded importance](https://elkronos.github.io/featR/articles/regularization.html)
 #'   guide, which runs this function on a worked example.
 #' @export

R/fs_recursivefeature.R

@@ -823,7 +823,6 @@ rfe_train_final <- function(data_df, target, optimal_vars,
 #' }
 #'
 #' @examples
-#' \donttest{
 #' if (requireNamespace("caret", quietly = TRUE) &&
 #'     requireNamespace("randomForest", quietly = TRUE) &&
 #'     requireNamespace("e1071", quietly = TRUE)) {
@@ -839,7 +838,6 @@ rfe_train_final <- function(data_df, target, optimal_vars,
 #'   # the only held-out estimate on the object
 #'   print(res$details$test_metrics)
 #' }
-#' }
 #' @seealso The [Wrapper methods](https://elkronos.github.io/featR/articles/wrappers.html)
 #'   guide, which runs this function on a worked example.
 #' @export

R/fs_stepwise.R

@@ -298,7 +298,6 @@ step_term_scores <- function(model, coefs, term_labels) {
 #' }
 #'
 #' @examples
-#' \donttest{
 #' if (requireNamespace("MASS", quietly = TRUE)) {
 #'   res <- fs_stepwise(mtcars, target = "mpg", direction = "both")
 #'   print(res$selected)
@@ -309,7 +308,6 @@ step_term_scores <- function(model, coefs, term_labels) {
 #'   refit <- stats::update(res$model, . ~ . - wt, data = mtcars)
 #'   print(stats::formula(refit))
 #' }
-#' }
 #' @seealso The [Wrapper methods](https://elkronos.github.io/featR/articles/wrappers.html)
 #'   guide, which runs this function on a worked example.
 #' @export

R/fs_svm.R

@@ -1166,7 +1166,6 @@ svm_performance <- function(predictions, actuals, task) {
 #'   \item{call}{The matched call.}
 #' }
 #' @examples
-#' \donttest{
 #' if (requireNamespace("caret", quietly = TRUE) &&
 #'     requireNamespace("kernlab", quietly = TRUE) &&
 #'     requireNamespace("e1071", quietly = TRUE)) {
@@ -1179,7 +1178,7 @@ svm_performance <- function(predictions, actuals, task) {
 #'     tune_grid = data.frame(C = 1),
 #'     seed = 42
 #'   )
-#'   res$details$performance
+#'   print(res$details$performance)
 #'
 #'   # SVM-RFE keeps the two most useful measurements
 #'   sel <- fs_svm(
@@ -1194,9 +1193,8 @@ svm_performance <- function(predictions, actuals, task) {
 #'     n_features = 2,
 #'     seed = 42
 #'   )
-#'   sel$selected
-#'   sel$scores
-#' }
+#'   print(sel$selected)
+#'   print(sel$scores)
 #' }
 #' @seealso The [Wrapper methods](https://elkronos.github.io/featR/articles/wrappers.html)
 #'   guide, which runs this function on a worked example.

cran-comments.md

@@ -4,46 +4,64 @@
 
 This is an update of featR from 0.1.0 to 0.2.0.
 
-It is a correctness release. Two statistical, methodological, and
-programmatic reviews of the whole package fixed silent wrong answers, hangs,
-crashes, and documentation that did not match behavior. Each fix has a
-regression test; the test suite grew from 2013 to 2560 passing assertions.
+It is a correctness release. A statistical, methodological, and programmatic
+review of the whole package fixed silent wrong answers, hangs, crashes, and
+documentation that did not match behavior. Each fix has a regression test.
 Some fixes change results or turn previously accepted, invalid input into an
 informative error. NEWS.md lists every change, with behavior changes first.
 
 ## Test environments
 
 * GitHub Actions (r-lib/actions): macOS latest (R release), Windows latest
   (R release), Ubuntu latest (R devel, R release, R oldrel-1).
-* Ubuntu (R release) with every suggested package except Boruta, bigstatsr,
-  and PRROC installed: 2560 passing assertions, 7 skipped for those packages.
+* win-builder (R devel).
+* Local Ubuntu (R release), `R CMD check --as-cran` twice: once with every
+  suggested package installed and `--run-donttest`, and once with no
+  suggested packages installed (`_R_CHECK_FORCE_SUGGESTS_=false`), to confirm
+  that examples, tests, and the vignette all run without them.
 
 ## R CMD check results
 
 0 errors | 0 warnings | 0 notes
 
+* Words that may be flagged as possibly misspelled in DESCRIPTION (Barnhill,
+  Boruta, Gabry, Guyon, Hastie, Kursa, Rudnicki, Tibshirani, Vapnik, Vehtari,
+  Zou) are author surnames from the cited references and the name of the
+  Boruta method.
+
 ## Reverse dependencies
 
 There are no reverse dependencies on CRAN.
 
 ## Notes for reviewers
 
+* As requested in the review of 0.1.0, Boruta is no longer in single quotes
+  in the Description (it names the method), and the methods are cited as
+  authors (year) <doi:...>.
 * All modeling engines (brms, caret, glmnet, randomForest, kernlab, earth,
   Boruta, and others) are in Suggests. Every function that needs one checks
-  availability with `requireNamespace()` and fails with an informative message
-  naming the packages to install. All examples that use a suggested package are
-  wrapped in `\donttest{}` behind `requireNamespace()` guards, and the test
-  suite skips those paths when the package is absent.
-* `fs_bayes()` compiles a Stan program per candidate model. Its example is in
-  `\donttest{}`, guarded on brms being installed, limited to two
-  single-predictor models, and wrapped in `try()`, so a machine without a
-  working Stan toolchain skips it rather than failing. The equivalent call is
-  exercised in the test suite under `skip_on_cran()`.
+  availability with `requireNamespace()` and stops with a message naming the
+  packages to install. Examples that use a suggested package are guarded with
+  `requireNamespace()`. The tests skip those paths when a package is absent.
+* Examples run in under 2 seconds each and are not wrapped in `\donttest{}`,
+  with two exceptions. `fs_bayes()` compiles a Stan program per candidate
+  model; its example is in `\donttest{}`, guarded on brms, limited to two
+  single-predictor models, and wrapped in `try()` so that a machine without a
+  working Stan toolchain skips it rather than failing. `fs_boruta()` is in
+  `\donttest{}` because its run time depends on the number of Boruta
+  iterations.
+* `fs_mars()` with `search = "random"` and a factor outcome attaches earth (a
+  suggested package) for the duration of that call with
+  `withr::local_package()`, because caret's random-grid code for earth looks
+  up `contr.earth.response` on the search path. On exit, earth and any
+  packages it attached are detached, so the user's search path is unchanged.
 * Parallel execution is opt-in throughout. Every function defaults to a single
-  worker, requested worker counts are capped at the detected core count, and no
-  example or test uses more than two cores.
-* The package never calls `set.seed()` unless the user supplies a `seed`
-  argument, and then restores the previous RNG state on exit via
-  `withr::local_seed()`.
-* No package code writes to the user's file system, home directory, working
-  directory, or global environment.
+  worker, requested worker counts are capped at the detected core count, and
+  no example or test uses more than two cores.
+* The package never seeds the random number generator unless the user
+  supplies a `seed` argument, and then restores the previous RNG state on exit
+  via `withr::local_seed()`.
+* Package code writes nothing to the user's home directory, working
+  directory, or global environment. The only file it creates is a temporary
+  backing file under `tempdir()` for `fs_pca()` on very large data, which is
+  removed when the call returns.