Revisit code comment from melodie switch

[hfrick]

Do we still need this code snippet and it should be integrated? Or can we delete it?

tune/R/grid_helpers.R

Lines 1 to 25 in 3544816

	# TODO old predict_model bits to check into
	# # TODO patch since parsnip does not record the column names when Surv objects
	# # are used with fit_xy()
	# if (model$spec$mode == "censored regression") {
	# model$preproc$y_var <- names(y_vals)
	# }
	#
	# if (length(orig_rows) != nrow(x_vals)) {
	# msg <- "Some assessment set rows are not available at prediction time."
	#
	# if (has_preprocessor_recipe(workflow)) {
	# msg <-
	# c(
	# msg,
	# i =
	# "Consider using {.code skip = TRUE} on any recipe steps that
	# remove rows to avoid calling them on the assessment set."
	#
	# )
	# } else {
	# msg <- c(msg, i = "Did your preprocessing steps filter or remove rows?")
	# }
	#
	# cli::cli_abort(msg)
	# }

[hfrick]

Sending Claude through the git log of tune and parsnip surfaced the following:

1. The patch about the column names is currently needed if we had an engine for a survival model with an XY interface (we do not, because we put a formula interface in front of glmnet in censored). I've opened xy_xy() sets y_var to character(0) for Surv objects parsnip#1355 to fix it there instead of patching here.

2. The check on the number of rows is something we lost when integrating postprocessing, we should bring it back. We should move it to the prepocessing stage, when we handle the assessment set, so that we can fail earlier.