Allow list columns?

[jonthegeek]

Minimal, reproducible example:

test_data <- tibble::tibble(
  a = list(mtcars, mtcars),
  b = 1:2
)
recipes::recipe(test_data, b ~ .)

Error in model.frame.default(formula, data) : 
  invalid type (list) for variable 'a'

I understand this might be a weird case, but there are steps in textrecipes that expect a list column, and I've written a step of my own that expects a tibble. I was trying to do some of the setup (which involves a java engine and didn't play nice with parallel) outside of the recipe, and then use that data in a couple different recipes, but recipes won't let me start with a list. I can understand this error if it comes out of the prep as a list, but I'm not done with it yet.

[topepo]

I think that we will eventually support this. I'll leave this open so that we can document thoughts/requirements here.

[kwiscion]

It turns out that recipes can handle list columns, but only when using vars and roles arguments instead of a formula.

test_data <- tibble::tibble(
  a = list(mtcars, mtcars),
  b = 1:2
)
recipes::recipe(test_data, vars = c("a", "b"), roles = c("outcome", "predictor"))

Data Recipe

Inputs:

      role #variables
   outcome          1
 predictor          1

The issue with formula originates from function stats::model.frame() used to extract predictors names in function recipes:::get_rhs_vars().

[EmilHvitfeldt]

An unexpected side-effect of #1283 means that recipes now support list-columns

library(recipes)

test_data <- tibble::tibble(
  a = list(mtcars, mtcars),
  b = 1:2
)

recipe(test_data, b ~ .)
#> 
#> ── Recipe ──────────────────────────────────────────────────────────────────────
#> 
#> ── Inputs
#> Number of variables by role
#> outcome:   1
#> predictor: 1

[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.