Issue #156: Feature collect_extracts

[jrosell]

I added collect_extracts method to the workflow_set objects.

As I run the tests, I seen some of them failing too. This is why I added some changes to test-fit_best.R and test-collect-notes.R too.

[simonpcouch]

Looking super strong! I definitely think you made the right move of starting with collect_notes() and mirroring the structure.

One hiccup here is that the .extracts column won't always be around unless the user supplies an extract function. If they don't, they'll see:

library(tidymodels)

set.seed(1)
folds <- rsample::vfold_cv(mtcars, v = 3)

wflow_set <-
   workflow_set(
      list(reg = mpg ~ ., nonlin = mpg ~ wt + 1 / sqrt(disp)),
      list(lm = parsnip::linear_reg())
   )

wflow_set_trained <-
   wflow_set %>%
   workflow_map("fit_resamples", resamples = folds)

collect_extracts(wflow_set_trained)
#> Error in `dplyr::mutate()`:
#> ℹ In argument: `extracts = list(collect_extracts(result))`.
#> ℹ In row 1.
#> Caused by error in `collect_extracts()`:
#> ! The `.extracts` column does not exist.
#> ℹ Please supply a control object (`?tune::control_grid()`) with a non-`NULL`
#>   `extract` argument during resample fitting.

^{Created on 2024-08-15 with reprex v2.1.1}

All of the dplyr::mutate() error context is a bit distracting / confusing, but I don't think it's worth the complexity of trying to catch and rethrow—let's stick with the approach you've proposed. That said, could you add a snapshot test for the case where the user doesn't supply an extract function but tries to collect_extracts()?

Also, could you add a NEWS entry advertising your changes?

[simonpcouch]

Awesome, thanks so much!