read_model_info_table() only once

[simonpcouch]

The read_model_info_table() helper is called 3 times per fit, and is read from file each time.

library(tidymodels)

bt <- bootstraps(mtcars, 100)

bm <- 
  bench::mark(
    total = 
      fit_resamples(linear_reg(), mpg ~ ., bt),
    read_model_info_table = 
      replicate(300, parsnip:::read_model_info_table()),
    check = FALSE
  )
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.

bm
#> # A tibble: 2 × 6
#>   expression                 min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr>            <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 total                    5.94s    5.94s     0.168      57MB     8.92
#> 2 read_model_info_table 215.87ms 219.19ms     4.57     25.1MB     3.04

As a percentage of total time, the check takes:

100 * as.numeric(bm$median[2]) / as.numeric(bm$median[1])
#> [1] 3.887301

This PR proposes, instead, assigning the function output directly into the namespace once and then referring to its output from then on.

^{Created on 2023-03-08 with reprex v2.0.2}

[EmilHvitfeldt]

Great find! 1 suggetion

[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.