Support passed through data?

[emstruong]

Hello,

Thank you for the package. I noticed that tidyrules:::tidy.rpart() seems to have trouble when the data is passed through apply as an additional argument as opposed to being a set argument. Is there a workaround? Sometimes I'd like to iterate through the data...

library(rpart)
library(tidyrules)
#> 
#> Attaching package: 'tidyrules'
#> The following object is masked from 'package:rpart':
#> 
#>     prune

x1 <- lapply(list(as.formula("Kyphosis ~ Age + Number + Start")), \(x) rpart(formula = x, data = kyphosis))
tidy(x1[[1]])
#> ---- Rulelist --------------------------------
#> ▶ Keys: NULL
#> ▶ Number of rules: 5
#> ▶ Model type: rpart
#> ▶ Estimation type: classification
#> ▶ Is validation data set: FALSE
#> 
#> 
#>   rule_nbr LHS                                    RHS   support confidence  lift
#>      <int> <chr>                                  <fct>   <int>      <dbl> <dbl>
#> 1        1 ( Start >= 8.5 ) & ( Start >= 14.5 )   abse…      29      0.968  1.22
#> 2        2 ( Start >= 8.5 ) & ( Start < 14.5 ) &… abse…      12      0.929  1.18
#> 3        3 ( Start >= 8.5 ) & ( Start < 14.5 ) &… abse…      14      0.812  1.03
#> 4        4 ( Start < 8.5 )                        pres…      19      0.571  2.72
#> 5        5 ( Start >= 8.5 ) & ( Start < 14.5 ) &… pres…       7      0.556  2.65
#> ----------------------------------------------


x2 <- lapply(list(as.formula("Kyphosis ~ Age + Number + Start")), rpart, data = kyphosis)
tidy(x2[[1]])
#> Error in is.data.frame(data): ..1 used in an incorrect context, no ... to look in

^{Created on 2024-11-18 with reprex v2.1.1}

[talegari]

@emstruong Thanks for raising the issue.
The issue seems to be with as.party method from partykit.

library(rpart)
library(tidyrules)
#> 
#> Attaching package: 'tidyrules'
#> The following object is masked from 'package:rpart':
#> 
#>     prune
x1 <- lapply(list(as.formula("Kyphosis ~ Age + Number + Start")), \(x) rpart(formula = x, data = kyphosis))
tidy(x1[[1]])
#> ---- Rulelist --------------------------------
#> ▶ Keys: NULL
#> ▶ Number of rules: 5
#> ▶ Model type: rpart
#> ▶ Estimation type: classification
#> ▶ Is validation data set: FALSE
#> 
#> 
#>   rule_nbr LHS                                    RHS   support confidence  lift
#>      <int> <chr>                                  <fct>   <int>      <dbl> <dbl>
#> 1        1 ( Start >= 8.5 ) & ( Start >= 14.5 )   abse…      29      0.968  1.22
#> 2        2 ( Start >= 8.5 ) & ( Start < 14.5 ) &… abse…      12      0.929  1.18
#> 3        3 ( Start >= 8.5 ) & ( Start < 14.5 ) &… abse…      14      0.812  1.03
#> 4        4 ( Start < 8.5 )                        pres…      19      0.571  2.72
#> 5        5 ( Start >= 8.5 ) & ( Start < 14.5 ) &… pres…       7      0.556  2.65
#> ----------------------------------------------
x2 <- lapply(list(as.formula("Kyphosis ~ Age + Number + Start")),
             rpart,
             data = kyphosis
             )

# this is know to fail
tidy(x2[[1]])
#> Error in is.data.frame(data): ..1 used in an incorrect context, no ... to look in

# error is originating from `partykit::as.party` call
partykit::as.party(x2[[1]])
#> Error in is.data.frame(data): ..1 used in an incorrect context, no ... to look in

We are using partykit to extract rules. For now, I will raise a bugreport in partykit. For long term, let me think if avoiding partykit makes sense.