Remove warnings for manually created enrollment and failure rates

[jdblischak]

Closes #316

I refactored the define_*_rate() and check_*_rate() functions so that the end user will no longer receive a warning message if they self-constructed enroll_rate and fail_rate. However, if they provide an invalid input, the error message recommends the appropriate define_*_rate() function

library("gsDesign2")

enroll_rate <- define_enroll_rate(
  duration = c(2, 2, 10),
  rate = c(3, 6, 9)
)

# new behavior: the column stratum was added by define_enroll_rate()
enroll_rate
## # A tibble: 3 × 3
## stratum duration  rate
## <chr>      <dbl> <dbl>
##   1 All            2     3
## 2 All            2     6
## 3 All           10     9

class(enroll_rate)
## [1] "enroll_rate" "tbl_df"      "tbl"         "data.frame" 
gs_design_ahr(enroll_rate = enroll_rate)

# remove the class
class(enroll_rate) <- class(enroll_rate)[-1]
class(enroll_rate)
## [1] "tbl_df"     "tbl"        "data.frame"

# new behavior: no warning for missing class
gs_design_ahr(enroll_rate = enroll_rate)

# New behavior: Recommend define_enroll_rate() in error message when invalid
# input is detected
enroll_rate_invalid <- enroll_rate[, -3]
gs_design_ahr(enroll_rate = enroll_rate_invalid)
## Error in check_enroll_rate(enroll_rate) : 
##   The variable `rate` can't be NULL. Try the helper function define_enroll_rate() to prepare valid input.

[LittleBeannie]

Hello @jdblischak , thank you for making the changes! I observed that the commit is forked in your personal GitHub repository. Could you please guide me on how to test the jdblischak:remove-input-class-warning branch? Should I clone it from your repository?

[jdblischak]

Two options for testing locally:

1. Use {remotes} in the R console

remotes::install_github("jdblischak/gsDesign2@remove-input-class-warning")

2. Use Git in the terminal

git remote add john https://github.com/jdblischak/gsDesign2.git
git fetch john
git checkout remove-input-class-warning
# Build and test the package in R

# When you are done, return to main branch
git checkout main

[nanxstats]

Now with this patch it's much clearer. Moving the validation logic to the validator check_*() and call the validator in the constructor define_*() makes this much better aligned with the patterns described in the S3 chapter of Advanced R.