Double intercept for preprocessing formula with dot notation in model formula

[hfrick]

Following on from tidymodels/censored#272, and moving beyond glmnet models, I found a bug in workflows where we end up with one too many intercepts...
Workflows adds an intercept in the pre stage, and then another gets added by the expansion of the . in the model formula.

library(parsnip)
library(workflows)

wflow_fit <- 
  workflow() %>% 
  add_formula(mpg ~ cyl + disp + hp) %>% 
  add_model(linear_reg(), formula = mpg ~ cyl + disp + hp) %>% 
  fit(data = mtcars)

# this is what we'd expect
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

wflow_fit <- 
  workflow() %>% 
  add_formula(mpg ~ cyl + disp + hp) %>% 
  add_model(linear_reg(), formula = mpg ~ .) %>% 
  fit(data = mtcars)

# this has one too many intercepts
coef(wflow_fit$fit$fit$fit)
#>   (Intercept) `(Intercept)`           cyl          disp            hp 
#>   34.18491917            NA   -1.22741994   -0.01883809   -0.01467933

^{Created on 2023-11-24 with reprex v2.0.2}

So far, I think this does only affect this specific case of preprocessing formula + dot in model formula.

Reprex for it not being the dot expansion in the preprocessing formula

library(parsnip)
library(workflows)

wflow_fit <- 
  workflow() %>% 
  add_formula(mpg ~ .) %>% 
  add_model(linear_reg(), formula = mpg ~ cyl + disp + hp) %>% 
  fit(data = mtcars)
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

^{Created on 2023-11-24 with reprex v2.0.2}

Reprex for it being the interaction with the preprocessing formula

library(parsnip)
library(workflows)
library(recipes)

# recipe as preprocessor --------------------------------------------------

wflow_fit <- 
  workflow() %>% 
  add_recipe(recipe(mpg ~ cyl + disp + hp, mtcars)) %>% 
  add_model(linear_reg(), formula = mpg ~ cyl + disp + hp) %>% 
  fit(data = mtcars)
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

wflow_fit <- 
  workflow() %>% 
  add_recipe(recipe(mpg ~ cyl + disp + hp, mtcars)) %>% 
  add_model(linear_reg(), formula = mpg ~ .) %>% 
  fit(data = mtcars)
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933


# variables as preprocessor -----------------------------------------------

wflow_fit <- 
  workflow() %>% 
  add_variables(outcomes = mpg, predictors = c(cyl, disp, hp)) %>% 
  add_model(linear_reg(), formula = mpg ~ cyl + disp + hp) %>% 
  fit(data = mtcars)
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

wflow_fit <- 
  workflow() %>% 
  add_variables(outcomes = mpg, predictors = c(cyl, disp, hp)) %>% 
  add_model(linear_reg(), formula = mpg ~ .) %>% 
  fit(data = mtcars)
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

^{Created on 2023-11-24 with reprex v2.0.2}

[hfrick]

Closed in tidymodels/parsnip#1033

library(parsnip)
library(workflows)

wflow_fit <- 
  workflow() %>% 
  add_formula(mpg ~ cyl + disp + hp) %>% 
  add_model(linear_reg(), formula = mpg ~ cyl + disp + hp) %>% 
  fit(data = mtcars)

# this is what we'd expect
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

wflow_fit <- 
  workflow() %>% 
  add_formula(mpg ~ cyl + disp + hp) %>% 
  add_model(linear_reg(), formula = mpg ~ .) %>% 
  fit(data = mtcars)

# this is okay now too
coef(wflow_fit$fit$fit$fit)
#> (Intercept)         cyl        disp          hp 
#> 34.18491917 -1.22741994 -0.01883809 -0.01467933

^{Created on 2023-12-11 with reprex v2.0.2}

[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.