Cannot retrieve original data object with "rpart" engine

[grantmcdermott]

The problem

Hi folks. I'm having trouble retrieving the original data object that was passed to the "rpart" engine. The old eval(tree$call$data) approach doesn't work here b/c the data object is hidden... and so it reverts back to the global utils::data() function.

I tried to see how you implement this as part of the update method for rpart backends. But unless I'm missing something, an _rpart.update method is not supported either.

Reproducible example

library(parsnip)
library(rpart)

tree = decision_tree() |>
  set_engine("rpart") |>
  set_mode("classification") |>
  fit(Species ~ Petal.Length + Petal.Width, data = iris)

tree$fit$call
#> rpart::rpart(formula = Species ~ Petal.Length + Petal.Width, 
#>     data = data)

# data object here is evaluated as the utils::data() function, not the actual passed data object
head(eval(tree$fit$call$data))
#>                                                                             
#> 1 function (..., list = character(), package = NULL, lib.loc = NULL,        
#> 2     verbose = getOption("verbose"), envir = .GlobalEnv, overwrite = TRUE) 
#> 3 {                                                                         
#> 4     fileExt <- function(x) {                                              
#> 5         db <- grepl("\\\\.[^.]+\\\\.(gz|bz2|xz)$", x)                     
#> 6         ans <- sub(".*\\\\.", "", x)

# update doesn't work either
update(tree)
#> Error in update.default(tree): need an object with call component

^{Created on 2024-07-15 with reprex v2.1.0}

Any suggestions for a workaround would be much appreciated. Thanks in advance.

P.S. Possibly related to https://github.com/tidymodels/parsnip/issues/257 but I need an automated way to retrieve the original data object. In this issue, the solution was to manually write it back to the call. Similarly, repair_call also needs the user to manually specify the data object.

[simonpcouch]

Thanks for the issue, @grantmcdermott!

Looks like rpart carries around the data it's fitted on as an attribute to its terms object:

library(parsnip)
library(rpart)

tree_parsnip <-
  decision_tree() |>
  set_engine("rpart") |>
  set_mode("classification") |>
  fit(Species ~ Petal.Length + Petal.Width, data = iris)

dim(attr(tree_parsnip$fit$terms, ".Environment")$data)
#> [1] 150   5
head(attr(tree_parsnip$fit$terms, ".Environment")$data)
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#> 1          5.1         3.5          1.4         0.2  setosa
#> 2          4.9         3.0          1.4         0.2  setosa
#> 3          4.7         3.2          1.3         0.2  setosa
#> 4          4.6         3.1          1.5         0.2  setosa
#> 5          5.0         3.6          1.4         0.2  setosa
#> 6          5.4         3.9          1.7         0.4  setosa

^{Created on 2024-07-15 with reprex v2.1.0}

If you're more intrigued by the on-label rpart::rpart(model) argument, which allows for explicitly opting to keep the model frame around, you can pass it as an "engine argument" by passing it in the usual form to set_engine():

library(parsnip)
library(rpart)

tree_parsnip <-
  decision_tree() |>
  set_engine("rpart", model = TRUE) |>
  set_mode("classification") |>
  fit(Species ~ Petal.Length + Petal.Width, data = iris)

dim(tree_parsnip$fit$model)
#> [1] 150   3
head(tree_parsnip$fit$model)
#>   Species Petal.Length Petal.Width
#> 1  setosa          1.4         0.2
#> 2  setosa          1.4         0.2
#> 3  setosa          1.3         0.2
#> 4  setosa          1.5         0.2
#> 5  setosa          1.4         0.2
#> 6  setosa          1.7         0.4

Note that the update() methods have methods for the model specifications rather than fitted models (e.g. decision_tree() rather than fit(decision_tree())), so there's no training data passed around there.

[grantmcdermott]

Ah, super. Thanks @simonpcouch! Much appreciated, bud.

[github-actions[bot]]

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.