Butcher on a glm with a data.frame that has a single column

[guiguilegui]

Hey!

The problem

I'm having trouble using butcher on a glm with a data.frame that has a single column:

Error in is_scalar_vector(x) && is.na(x) :
'length = 10' in coercion to 'logical(1)'

Digging a little bit, I see it's a problem using rlang::is_na(out) from the butcher:::exchange function.

Reproducible example

library(butcher)
#This does not work
df <- data.frame(y = 1:10)
model <- glm(y ~1, data = df)
butcher(model)

#Adding a second column is a workaround
df <- data.frame(y = 1:10, x = 1:10)
model <- glm(y ~1, data = df)
butcher(model)

#For comparison, a single column works with lm
df <- data.frame(y = 1:10)
model <- lm(y ~ 1, data = df)
butcher(model)

#Session Info:

sessionInfo()
# R version 4.4.2 (2024-10-31 ucrt)                                                                        
# Platform: x86_64-w64-mingw32/x64                                                                         
# Running under: Windows 11 x64 (build 22631)                                                              
                                                                                                         
# Matrix products: default                                                                                 
                                                                                                         
                                                                                                         
# locale:
# [1] LC_COLLATE=French_Canada.1252  LC_CTYPE=French_Canada.1252    LC_MONETARY=French_Canada.1252
# [4] LC_NUMERIC=C                   LC_TIME=French_Canada.1252

# time zone: America/Toronto
# tzcode source: internal

# attached base packages:
# [1] stats     graphics  grDevices datasets  utils     methods   base

# other attached packages:
# [1] butcher_0.3.4

# loaded via a namespace (and not attached):
# [1] compiler_4.4.2 tools_4.4.2    renv_1.0.11

Thanks for the great package!

[juliasilge]

I am not quite sure why we are using rlang::is_na() in this way:

butcher/R/utilities.R

Line 5 in 0d3bf50

if (!rlang::is_na(out)[1]) {

It says it is always supposed to have a scalar predicate and return a scalar, but notice how we have [1] there like we are trying to subset.

df <- data.frame(y = 1:10)
model <- glm(y ~1, data = df)
out <- purrr::pluck(model, "data", .default = NA)
rlang::is_na(out)
#> Error in is_scalar_vector(x) && is.na(x): 'length = 10' in coercion to 'logical(1)'

^{Created on 2025-01-10 with reprex v2.1.1}

Looks like back in 2019 we were trying to solve #88. Maybe instead we should use NULL, like this:

df <- data.frame(y = 1:10)
mod <- glm(y ~ 1, data = df)
out <- purrr::pluck(mod, "data", .default = NULL)
rlang::is_null(out)
#> [1] FALSE


mod <- earth::earth(Volume ~ ., data = trees)
out <- purrr::pluck(mod, "data", .default = NULL)
rlang::is_null(out)
#> [1] TRUE

^{Created on 2025-01-10 with reprex v2.1.1}

[juliasilge]

Thank you again for this report @guiguilegui! I have fixed this problem in #278 and you can install the version of butcher with the fix via:

# install.packages("pak")
pak::pak("tidymodels/butcher")