time to eval_time

[hfrick]

Closes #241

This PR completes the deprecation of the time argument to the prediction functions in favor of the eval_time argument name laid out in tidymodels/parsnip#892.

The commits are (mostly) working through the package model/engine combination by model/engine combination, but the aggregated pattern is

• In the -data.R files, prediction modules for types "survival" and "hazard" need to expect the eval_time arg instead of the time arg
• The censored functions to calculate survival or hazard probabilities from a model object, e.g., survival_prob_coxph(), also make the switch and since they are exported, time is being deprecated. Since they are also @keywords internal, there are no specific deprecation tests. Happy to add those but thought that this might be a bit much.
• The deprecation "level" is "warn" as in time to eval_time parsnip#936
• The internal helper functions also now use eval_time where appropriate.
• The tests for the deprecation of the time arg in parsnip::predict_survival() and parsnip::predict_hazard() are in test-survival_reg-survival.R

[hfrick]

@simonpcouch The changes are pretty boring to read - the more challenging aspect is if I overlooked anything to add. If you could tilt your review toward that, that'd be great! 🙌

[simonpcouch]

I’m on board with your deprecation approach—thumbs up for the unit testing comment and transition immediately to the more aggressive deprecate_warn().

I’m guessing many of these are false positives, but a few places where .time or time is still around in a way that isn’t immediately obvious to me:

censored/R/aaa_survival_prob.R

Line 28 in e61368d

.time = x$time,

censored/R/aaa_survival_prob.R

Line 44 in e61368d

.time = rep(x$time, n),

censored/R/aaa_survival_prob.R

Line 57 in e61368d

.time = 0,

censored/R/boost_tree-mboost.R

Lines 192 to 193 in b8305bb

	next_event_time = dplyr::lead(.time),
	time_interval = next_event_time - .time,

censored/R/boost_tree-mboost.R

Line 124 in e61368d

# survival_prob is length(time) x nrow(new_data) and

I believe the intent here was to test formula vs xy rather than with the two different arguments:

censored/tests/testthat/test-survival_reg-survival.R

Line 200 in b8305bb

time = c(100, 200)

There are a good few places in this vignette that use time:

censored/vignettes/articles/examples.Rmd

Line 62 in b8305bb

time = c(100, 500, 1000)

[simonpcouch]

Note to self / future GitHub sleuth--times is the argument to survival::summary.survfit()

[hfrick]

Thank you! The .time vectors are indeed false positives. I plan on changing those when I make up my mind between calling them observed times or event times. The other ones were very helpful, thanks!

[hfrick]

CI failures are due to changes in the new pillar, so unrelated to this PR

[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.