Add support for updating recipe/model in workflowset object

[juliasilge]

We would like to support the functionality of update_recipe() and update_model() for the workflowset object, so for example someone can use a sparse blueprint or a special model formula.

I do find it worrisome that this currently runs without error, even though it is not doing anything with sparse_bp:

library(tidymodels)
#> Registered S3 method overwritten by 'tune':
#>   method                   from   
#>   required_pkgs.model_spec parsnip

data(parabolic)
set.seed(1)
split <- initial_split(parabolic)
train_set <- training(split)
test_set <- testing(split)
set.seed(2)
train_resamples <- bootstraps(train_set)


glmnet_spec <- 
  logistic_reg(penalty = 0.1, mixture = 0) %>%
  set_engine("glmnet")

rec <-
  recipe(class ~ ., data = train_set) %>%
  step_YeoJohnson(all_numeric_predictors())

sparse_bp <- hardhat::default_recipe_blueprint(composition = "dgCMatrix")

workflow_set(
  preproc = list(yj = rec, 
                 norm = rec %>% step_normalize(all_numeric_predictors())),
  models = list(regularized = glmnet_spec)
) %>%
  option_add(blueprint = sparse_bp) %>%
  workflow_map(fn = "fit_resamples", resamples = train_resamples)
#> # A workflow set/tibble: 2 × 4
#>   wflow_id         info             option    result   
#>   <chr>            <list>           <list>    <list>   
#> 1 yj_regularized   <tibble [1 × 4]> <opts[2]> <rsmp[+]>
#> 2 norm_regularized <tibble [1 × 4]> <opts[2]> <rsmp[+]>

^{Created on 2021-12-09 by the reprex package (v2.0.1)}

IMO there are three things to address:

• support the update_recipe() / update_model() functionality
• possibly error or warn if something is passed to option_add() that does nothing?
• document (if it is already here) or create a way to get at the fitted workflow result, so I could for example call extract_mold() and see what happened during fitting (maybe just the extract arg of the control object?)

[topepo]

Do either of these patterns seem preferable to you or @DavisVaughan?

# ------------------------------------------------------------------------------
# option 1

update_workflow <-
  function(x, id, recipe = NULL,  blueprint = NULL, spec = NULL, formula = NULL) {
    
    wflow <- extract_workflow(x, id = id)
    
    # update_recipe and/or update_model
    
    # stuff back into workflow set
    
  }

# ------------------------------------------------------------------------------
# option 2

update_workflow_model <- function(x, id, spec, formula = NULL) {
    # stuff
}

update_workflow_recipe <- function(x, id, recipe, blueprint = NULL) {
  # stuff
}

[juliasilge]

I think I prefer option 2 but that is a weakly held opinion.

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.