code style in checking internals

[qiushiyan]

When mode == 'unknown', inform_missing_implementation searches available models for all modes.

Also had to use new_model_spec(check_missing_spec = FALSE) in boost_tree (and perhaps other models that involve extension packages) and new_model_spec(check_missing_spec = TRUE) in set_engine.model_spec(), such that inform doesn't use the default engine

# before: parsnip could not locate an implementation for bag_tree model specifications using the `rpart` engine ... 
boost_tree() %>% 
  set_engine("C5.0")

Current error messages

library(parsnip)

bag_tree() %>%
  set_engine("rpart") %>%
  testthat::capture_message()
#> <message/rlang_message>
#> Message:
#> ℹ parsnip could not locate an implementation for bag_tree model specifications using the `rpart` engine. The following parsnip extension packages implement support for this specification.
#> ℹ censored, baguette
#> ℹ Please install one of them (if needed) and load to continue.

bag_tree() %>% 
  set_mode("censored regression") %>% 
  set_engine("rpart") %>% 
  testthat::capture_message()
#> <message/rlang_message>
#> Message:
#> ℹ parsnip could not locate an implementation for bag_tree censored regression model specifications using the `rpart` engine. Please install `censored` (if needed) and load to continue.

bag_tree() %>% 
  set_engine("C5.0") %>% 
  testthat::capture_message()
#> <message/rlang_message>
#> Message:
#> ℹ parsnip could not locate an implementation for bag_tree model specifications using the `C5.0` engine. Please install `baguette` (if needed) and load to continue.

library(censored)
#> Loading required package: survival

bag_tree() %>% 
  set_mode("censored regression") %>% 
  set_engine("rpart") %>% 
  testthat::capture_message()
#> NULL

library(baguette)
bag_tree() %>% 
  set_engine("C5.0") %>% 
  testthat::capture_message()
#> NULL

^{Created on 2022-08-01 by the reprex package (v2.0.1)}
One remaining problem being if set_mode is called after set_engine then the user specified mode won't be reflected in the error message. If set_mode uses new_model_spec(check_missing_arg = TRUE) as well then we get two sets of messages.

[simonpcouch]

Thanks for making this happen, @qiushiyan! Feel free to mark this PR as ready for review when you feel good about it and I'll take a look. I had wondered if there were any conditions that this might happen when I put this together, and it looks like I might have missed one. The argument checking across model spec functions is indeed tricky. :)

[qiushiyan]

Now compute inform message at print time in print_model_spec() with davis suggestions 🎉 . I think changes in fit() need more thoughts could be in a separate PR. @simonpcouch could you take a look and take this over?

[simonpcouch]

Awesome, thanks for making this happen @qiushiyan. Will review and move this along next week. :)

[simonpcouch]

Mostly reviewing for myself to point out notable changes.

Lots of helpful syntax consistency changes here, though the diffs are a bit difficult to navigate because of them.

Otherwise, as is, relocates check_missing_spec from new_model_spec to print.model_spec, shortens the missing extension package message, and makes needed adjustments!

From here, as Qiushi mentioned, need to figure out actions that indicate defining a model spec is "finished" so we know we can check the mode vs engine vs class. fit is one of those places.

[simonpcouch]

Splitting this PR into two, one for code style edits and one to improve messaging for missing implementations. This one is now the former, so will go ahead and merge!

[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.