More metric summarizer

[EmilHvitfeldt]

This PR adds a refactoring of metric_summarizer() by splitting it into 3 new functions metric_summarizer_class(), metric_summarizer_numeric(), and metric_summarizer_prob(). (metric_summarizer() is being soft deprecated)

This will lessen the burden of modifying any one of the types of metrics as metric_summarizer() was forced to cover all cases. This PR is one step towards adding survival metrics.

metric_summarizer_class() and metric_summarizer_prob() are carbon copies of metric_summarizer(), where on the other hand , metric_summarizer_numeric() doesn't have an estimator or event_level arguments as they aren't used for numeric metrics.

I decided to document them separately, as I found it less confusing to have for numeric metrics this is NULL, for class metrics ....

[EmilHvitfeldt]

I have not contacted revdep authors yet. Will do that before merging

[DavisVaughan]

Ok I've given a first round of review. I am fully expecting this to be a rather large PR with a few iterations. Feel free to ping me with questions, and when you finish this round

[DavisVaughan]

I've got some ideas to make truth, estimate, case_weights, and ... use tidyselect rather than allowing arbitrary expressions through here (which in retrospect is a bit horrifying). We can talk about it after this first round of review, but I think it will simplify these functions a bit further.

If I'm right, we will be able to do something like .data[[truth]] rather than !! which would be nice

[DavisVaughan]

I think that will also automatically support this special case so we would be able to remove

  # Explicit handling of length 1 character vectors as column names
  nms <- colnames(data)
  truth <- handle_chr_names(truth, nms)
  estimate <- handle_chr_names(estimate, nms)

That would work automatically because tidyselection allows strings as well as bare expressions

[DavisVaughan]

I'm hoping to remove all of the spliceable_case_weights() style helpers too, which we can also talk about

[DavisVaughan]

I think this is looking good! One more round of review. In the meantime I'll also think about the tidyselect bits more and send you a message about it.

[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.