add toggle_sparsity()

DESCRIPTION

@@ -25,6 +25,7 @@ Imports:
     lifecycle (>= 1.0.3),
     modelenv (>= 0.1.0),
     parsnip (>= 1.2.1.9000),
+    recipes (>= 1.0.10.9000),
     rlang (>= 1.1.0),
     tidyselect (>= 1.2.0),
     sparsevctrs (>= 0.1.0.9002),
@@ -42,7 +43,6 @@ Suggests:
     methods,
     modeldata (>= 1.0.0),
     probably,
-    recipes (>= 1.0.10.9000),
     rmarkdown,
     testthat (>= 3.0.0)
 VignetteBuilder: 

R/fit.R

@@ -71,6 +71,8 @@ fit.workflow <- function(object, data, ..., calibration = NULL, control = contro
     )
   }
 
+  object <- toggle_sparsity(object, data)
+
   workflow <- object
   workflow <- .fit_pre(workflow, data)
   workflow <- .fit_model(workflow, control)

R/sparsevctrs.R

@@ -1,3 +1,79 @@
 is_sparse_matrix <- function(x) {
   methods::is(x, "sparseMatrix")
 }
+
+toggle_sparsity <- function(object, data) {
+  toggle_sparse <- "no"
+
+  if (allow_sparse(object$fit$actions$model$spec)) {
+    if ("recipe" %in% names(object$pre$actions)) {
+      est_sparsity <- recipes::.recipes_estimate_sparsity(
+        object$pre$actions$recipe$recipe
+      )
+    } else {
+      est_sparsity <- sparsevctrs::sparsity(data, 1000)
+    }
+
+    pred_log_fold <- pred_log_fold(
+      est_sparsity,
+      object$fit$actions$model$spec$engine,
+      nrow(data)
+    )
+    if (pred_log_fold > 0) {
+      toggle_sparse <- "yes"
+    }
+  }
+
+  object$pre$actions$recipe$recipe <- recipes::.recipes_toggle_sparse_args(
+    object$pre$actions$recipe$recipe,
+    choice = toggle_sparse
+  )
+  object
+}
+
+allow_sparse <- function(x) {
+  if (inherits(x, "model_fit")) {
+    x <- x$spec
+  }
+  res <- parsnip::get_from_env(paste0(class(x)[1], "_encoding"))
+  all(res$allow_sparse_x[res$engine == x$engine])
+}
+
+pred_log_fold <- function(sparsity, model, n_rows) {
+  if (is.null(model) || model == "ranger") {
+    return("no")
+  }
+
+  log_fold <- -0.599333138645995 +
+    ifelse(sparsity < 0.836601307189543, 0.836601307189543 - sparsity, 0) *
+      -0.541581853008009 +
+    ifelse(n_rows < 16000, 16000 - n_rows, 0) * 3.23980908942813e-05 +
+    ifelse(n_rows > 16000, n_rows - 16000, 0) * -2.81001152147355e-06 +
+    ifelse(sparsity > 0.836601307189543, sparsity - 0.836601307189543, 0) *
+      9.82444255114058 +
+    ifelse(sparsity > 0.836601307189543, sparsity - 0.836601307189543, 0) *
+      ifelse(n_rows > 8000, n_rows - 8000, 0) *
+      7.27456967763306e-05 +
+    ifelse(sparsity > 0.836601307189543, sparsity - 0.836601307189543, 0) *
+      ifelse(n_rows < 8000, 8000 - n_rows, 0) *
+      -0.000798307404212627
+
+  if (model == "xgboost") {
+    log_fold <- log_fold +
+      ifelse(sparsity < 0.984615384615385, 0.984615384615385 - sparsity, 0) *
+        0.113098025073806 +
+      ifelse(n_rows < 8000, 8000 - n_rows, 0) * -9.77914237255269e-05 +
+      ifelse(n_rows > 8000, n_rows - 8000, 0) * 3.22657666511869e-06 +
+      ifelse(sparsity > 0.984615384615385, sparsity - 0.984615384615385, 0) *
+        41.5180348086939 +
+      0.913457808326756
+  }
+
+  if (model == "LiblineaR") {
+    log_fold <- log_fold +
+      ifelse(sparsity > 0.836601307189543, sparsity - 0.836601307189543, 0) *
+        -5.39592564852111
+  }
+
+  log_fold
+}