Why is "preproc" designated by the type rather than the name of the preprocessor?

[joranE]

Perhaps I'm missing something obvious, but why is the preproc field in the info column populated with the preprocessor type rather than the name, as passed in when calling workflow_set?

My expectation would have been that autoplot would distinguish between different preprocessors using shape, but that appears to only happen if they are passed in as different types:

library(workflowsets)
library(workflows)
library(modeldata)
library(recipes)
library(parsnip)
library(dplyr)
library(rsample)
library(tune)
library(yardstick)

# ------------------------------------------------------------------------------

data(two_class_dat, package = "modeldata")

set.seed(1)
folds <- vfold_cv(two_class_dat, v = 5)

# ------------------------------------------------------------------------------

decision_tree_rpart_spec <-
  decision_tree(min_n = tune(), cost_complexity = tune()) |>
  set_engine('rpart') |>
  set_mode('classification')

logistic_reg_glm_spec <-
  logistic_reg() |>
  set_engine('glm')

mars_earth_spec <-
  mars(prod_degree = tune()) |>
  set_engine('earth') |>
  set_mode('classification')

# ------------------------------------------------------------------------------

def_recipe <- 
  recipe(Class ~ A + B,data = two_class_dat)

yj_recipe <-
  recipe(Class ~ ., data = two_class_dat) |>
  step_YeoJohnson(A, B)

# ------------------------------------------------------------------------------

two_class_set1 <-
  workflow_set(
    preproc = list(none = Class ~ A + B, yj_trans = yj_recipe),
    models = list(cart = decision_tree_rpart_spec, glm = logistic_reg_glm_spec,
                  mars = mars_earth_spec)
  )

two_class_set2 <-
  workflow_set(
    preproc = list(none = def_recipe, yj_trans = yj_recipe),
    models = list(cart = decision_tree_rpart_spec, glm = logistic_reg_glm_spec,
                  mars = mars_earth_spec)
  )


# ------------------------------------------------------------------------------

two_class_res1 <-
  two_class_set1 |>
  workflow_map(
    resamples = folds,
    grid = 10,
    seed = 2,
    verbose = TRUE,
    control = control_grid(save_workflow = TRUE)
  )

two_class_res2 <-
  two_class_set2 |>
  workflow_map(
    resamples = folds,
    grid = 10,
    seed = 2,
    verbose = TRUE,
    control = control_grid(save_workflow = TRUE)
  )

# Uses shape for the two different preprocessors
autoplot(two_class_res1)

# All preprocessors are simply "recipe" and you can't tell them apart
autoplot(two_class_res2)

[topepo]

Because the number of workflow IDs could be very large, beyond the number of possible symbols that are available.

I'm open to an alternative autoplot() method so long as it can handle the ggplot constraints.