Reorder columns at prediction time for glmnet

[juliasilge]

Closes #273.

I played around in the various predict functions and methods today, and I think the best place to do this reordering is in the prediction module for the two glmnet models. This is mainly because:

• the fitted object is available
• all the preprocessing has already been done

library(parsnip)
library(tidyverse)
set.seed(989)
# Gaussian
x = matrix(rnorm(100 * length(letters)), 100, length(letters)) %>%
  magrittr::set_colnames(letters)
y = rnorm(100)

glmnet_spec <- linear_reg(penalty = .1) %>%
  set_engine('glmnet')

glmnet_fit <- fit_xy(glmnet_spec, x, y)

glmnet_fit %>%
  predict(x)
#> # A tibble: 100 x 1
#>      .pred
#>      <dbl>
#>  1 -0.135 
#>  2  0.285 
#>  3 -0.126 
#>  4  0.222 
#>  5 -0.138 
#>  6 -0.157 
#>  7  0.0271
#>  8 -0.0215
#>  9 -0.0702
#> 10 -0.0283
#> # … with 90 more rows

glmnet_fit %>%
  predict(x[, sample(colnames(x))])
#> # A tibble: 100 x 1
#>      .pred
#>      <dbl>
#>  1 -0.135 
#>  2  0.285 
#>  3 -0.126 
#>  4  0.222 
#>  5 -0.138 
#>  6 -0.157 
#>  7  0.0271
#>  8 -0.0215
#>  9 -0.0702
#> 10 -0.0283
#> # … with 90 more rows



## still works with dummy variables
data(ames, package = "modeldata")

parsnip_form_fit <- glmnet_spec %>%
  fit(Sale_Price ~ Year_Built + Alley, data = ames)

predict(parsnip_form_fit, ames %>% select(Year_Built, Alley))
#> # A tibble: 2,930 x 1
#>      .pred
#>      <dbl>
#>  1 163033.
#>  2 164552.
#>  3 159996.
#>  4 175180.
#>  5 219215.
#>  6 220733.
#>  7 225289.
#>  8 211623.
#>  9 216178.
#> 10 222252.
#> # … with 2,920 more rows

^{Created on 2020-10-20 by the reprex package (v0.3.0.9001)}

I ran extratests with this change and there were no failures related to this.

[topepo]

This needs to happen in the poissonreg package too.

[github-actions]

This pull request 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.