switch to fit_xy() if sparse matrix is passed to fit()

Author: EmilHvitfeldt

R/fit.R

@@ -175,10 +175,13 @@ fit.model_spec <-
     eval_env$weights <- wts
 
     if (is_sparse_matrix(data)) {
-      cli::cli_abort(c(
-        x = "Sparse matrices cannot be used with {.fn fit}.",
-        i = "Please use {.fn fit_xy} interface instead."
-      ))
+      outcome_names <- all.names(rlang::f_lhs(formula))
+      outcome_ind <- match(outcome_names, colnames(data))
+
+      y <- data[, outcome_ind]
+      x <- data[, -outcome_ind, drop = TRUE]
+
+      return(fit_xy(object, x, y, case_weights, control, ...))
     }
 
     data <- materialize_sparse_tibble(data, object, "data")

tests/testthat/_snaps/sparsevctrs.md

@@ -6,6 +6,14 @@
       Warning:
       `data` is a sparse tibble, but `linear_reg()` with engine "lm" doesn't accept that. Converting to non-sparse.
 
+# sparse matrix can be passed to `fit()
+
+    Code
+      lm_fit <- fit(spec, avg_price_per_room ~ ., data = hotel_data[1:100, ])
+    Condition
+      Error in `fit_xy()`:
+      ! `x` is a sparse matrix, but `linear_reg()` with engine "lm" doesn't accept that.
+
 # sparse tibble can be passed to `fit_xy()
 
     Code
@@ -22,15 +30,6 @@
       Error in `fit_xy()`:
       ! `x` is a sparse matrix, but `linear_reg()` with engine "lm" doesn't accept that.
 
-# sparse matrices can not be passed to `fit()
-
-    Code
-      hotel_fit <- fit(spec, avg_price_per_room ~ ., data = hotel_data)
-    Condition
-      Error in `fit()`:
-      x Sparse matrices cannot be used with `fit()`.
-      i Please use `fit_xy()` interface instead.
-
 # sparse tibble can be passed to `predict()
 
     Code

tests/testthat/test-sparsevctrs.R

@@ -21,33 +21,34 @@ test_that("sparse tibble can be passed to `fit()", {
   )
 })
 
-test_that("sparse tibble can be passed to `fit_xy()", {
+test_that("sparse matrix can be passed to `fit()", {
   skip_if_not_installed("xgboost")
 
   hotel_data <- sparse_hotel_rates()
-  hotel_data <- sparsevctrs::coerce_to_sparse_tibble(hotel_data)
 
   spec <- boost_tree() %>%
     set_mode("regression") %>%
     set_engine("xgboost")
 
   expect_no_error(
-    lm_fit <- fit_xy(spec, x = hotel_data[, -1], y = hotel_data[, 1])
+    lm_fit <- fit(spec, avg_price_per_room ~ ., data = hotel_data)
   )
 
   spec <- linear_reg() %>%
     set_mode("regression") %>%
     set_engine("lm")
 
   expect_snapshot(
-    lm_fit <- fit_xy(spec, x = hotel_data[1:100, -1], y = hotel_data[1:100, 1])
+    error = TRUE,
+    lm_fit <- fit(spec, avg_price_per_room ~ ., data = hotel_data[1:100, ])
   )
 })
 
-test_that("sparse matrices can be passed to `fit_xy()", {
+test_that("sparse tibble can be passed to `fit_xy()", {
   skip_if_not_installed("xgboost")
 
   hotel_data <- sparse_hotel_rates()
+  hotel_data <- sparsevctrs::coerce_to_sparse_tibble(hotel_data)
 
   spec <- boost_tree() %>%
     set_mode("regression") %>%
@@ -62,12 +63,11 @@ test_that("sparse matrices can be passed to `fit_xy()", {
     set_engine("lm")
 
   expect_snapshot(
-    lm_fit <- fit_xy(spec, x = hotel_data[1:100, -1], y = hotel_data[1:100, 1]),
-    error = TRUE
+    lm_fit <- fit_xy(spec, x = hotel_data[1:100, -1], y = hotel_data[1:100, 1])
   )
 })
 
-test_that("sparse matrices can not be passed to `fit()", {
+test_that("sparse matrices can be passed to `fit_xy()", {
   skip_if_not_installed("xgboost")
 
   hotel_data <- sparse_hotel_rates()
@@ -76,9 +76,17 @@ test_that("sparse matrices can not be passed to `fit()", {
     set_mode("regression") %>%
     set_engine("xgboost")
 
+  expect_no_error(
+    lm_fit <- fit_xy(spec, x = hotel_data[, -1], y = hotel_data[, 1])
+  )
+
+  spec <- linear_reg() %>%
+    set_mode("regression") %>%
+    set_engine("lm")
+
   expect_snapshot(
-    error = TRUE,
-    hotel_fit <- fit(spec, avg_price_per_room ~ ., data = hotel_data)
+    lm_fit <- fit_xy(spec, x = hotel_data[1:100, -1], y = hotel_data[1:100, 1]),
+    error = TRUE
   )
 })