add a metric print method

[simonpcouch]

Closes #435!

[EmilHvitfeldt]

As a general comment, i think we should overhaul the printing of metric_set() at the same time 😄

[simonpcouch]

Nice! Related to #454, then. How would you like to move this PR forward/not?

I'd be glad to work on the metric set method, too. I guess a one-line-output metric print method would make the metric set print method nice and consistent. :)

[EmilHvitfeldt]

I think we should add the metric set printing to this PR. we might learn things when updating that, that informs the individual printing method, so keeping them together seems easier

[simonpcouch]

I'm thinking this could look like:

roc_auc
#> A probability metric | direction: maximize

demographic_parity(boop)
#> A class metric       | direction: minimize, group-wise on: boop

...with:

metric_set(roc_auc, demographic_parity(boop))
#> A metric set, consisting of:
#> - A probability metric | direction: maximize
#> - A class metric       | direction: minimize, group-wise on: boop

...with cli formatting where needed. Notably, I think it'd be great to link probability metric and class metric to a {.help}-page clarifying what that means, if possible.

Thoughts?

[EmilHvitfeldt]

I like the look of that!

I would like to have the names of the metrics in the metric set printing.

[simonpcouch]

Spent some time on this yesterday but didn't quite have time to finish—moving on to other things for at least today. Still need to write code for justification (see rlib/cli#229).

[simonpcouch]

Currently, output looks like:

library(yardstick)
dem_parity_boop <- demographic_parity(boop)

roc_auc 
#> A probability metric | direction: maximize

dem_parity_boop
#> A class metric | direction: minimize, group-wise on: boop

demographic_parity(boop)
#> A class metric | direction: minimize, group-wise on: boop

metric_set(roc_auc, dem_parity_boop)
#> A metric set, consisting of:
#> - `roc_auc()`, a probability metric   | direction: maximize
#> - `dem_parity_boop()`, a class metric | direction: minimize, group-wise on:
#> boop

metric_set(roc_auc, demographic_parity(boop))
#> A metric set, consisting of:
#> - `roc_auc()`, a probability metric            | direction: maximize
#> - `demographic_parity(boop)()`, a class metric | direction: minimize,
#> group-wise on: boop

^{Created on 2023-11-27 with reprex v2.0.2}

While format.metric_set() does call format.metric, there's a good bit of post-processing that needs to happen to make justification work. From my reading of cli issues, we don't have many justification features to draw from there. I generally do feel like the rows x columns appearance for printing works well for metric sets and miss that structure with my current attempt. I may hand this off to you here, @EmilHvitfeldt, either for review or revision.

[EmilHvitfeldt]

This all looks good! added a couple of small changes, will merge when CI runs clean

[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.