Check compatibility of model spec and tailor

[hfrick]

closes #291

Parsnip model_specs can have a mode of "classification", "regression", "censored regression" or "unknown". When adding a model spec to a workflow, the mode cannot be unknown anymore.

Tailors can have a type of "regression", "binary", "multiclass" or "unknown". There is currently no way for the user to set the type directly, so workflows does not require the tailor to have a type other than unknown.

Tailors start with an unknown mode (via tailor::tailor()) and then the type gets inferred from the adjustments added.

When we add a tailor to a workflow, it will have

• type = "unknown" only if it's a tailor without any adjustments, or all adjustments are adjust_predictions_custom().
• type = "binary" for all classification problems, including multi-class. The type only gets updated during fit.tailor().
• or otherwise: type = "regression"

Once it comes to fitting a tailor, we have the response, so we can update the tailor type

• from "unknown" to one of the other ones
• from "binary" to "multiclass" if appropriate

This PR checks what we can when specifying the workflow, in add_tailor(), add_model(), and workflows().

In those situations where fit.tailor() changes the tailor type, we can't really do anything earlier, or are fine already:

• If the "unknown" tailor type gets changed, it was either a plain tailor (which is fine) or only custom adjustments (which we can't check).
• If the type gets changed from "binary" to "multiclass", we already checked that the model mode is "classification".

[Copilot]

Pull Request Overview

This PR adds compatibility validation between parsnip model specifications and tailor postprocessors in workflows. The validation ensures that model modes (regression, classification, etc.) are compatible with tailor types (regression, binary, multiclass) before adding them to a workflow.

• Adds compatibility checks in add_tailor(), add_model(), and workflow() functions
• Implements validation logic to prevent incompatible model-tailor combinations
• Adds comprehensive test coverage for the new validation functionality

Reviewed Changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
R/workflow.R	Updates error message format and adds validation call to add_postprocessor()
R/post-action-tailor.R	Implements core compatibility validation logic between model specs and tailors
R/fit-action-model.R	Adds validation when adding models to workflows with existing tailors
tests/testthat/test-workflow.R	Tests compatibility validation and postprocessor validation in main workflow function
tests/testthat/test-post-action-tailor.R	Tests compatibility validation when adding tailors to workflows
tests/testthat/test-fit-action-model.R	Tests compatibility validation when adding models to workflows with tailors
tests/testthat/test-extract.R	Updates test to use compatible model type for existing tailor
tests/testthat/_snaps/*.md	Snapshot files for error message testing

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[simonpcouch]

So, so helpful to have that description as a refresher. Thank you for the thoroughness!

Looks great.

[simonpcouch]

super helpful comment

[simonpcouch]

tbh finding it very encouraging that this successfully surfaced an issue in tests

[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.