refactor: use checkmate for arg validation

[dshemetov]

Related to cmu-delphi/epiprocess#403.

I took the safe approach of changing the internals of the arg functions to use checkmate, that way we can rely on the existing unit tests and don't need to change any other epipredict code. The functions look a lot simpler and more unified now. We could eventually get rid of these wrappers and just invoke check functions directly, but that would take more refactoring work.

[dshemetov]

Dates now pass arg_is_numeric, this didn't affect the rest of the tests.

[dshemetov]

Dates now pass arg_is_positive, this didn't affect the rest of the tests.

[dshemetov]

Dates now pass arg_is_nonneg, this didn't affect the rest of the tests.

[dshemetov]

I had to remove allow_na from arg_is_date, due to a checkmate inconsistency, but this didn't affect the rest of the tests.

[dshemetov]

These weren't used anywhere and checkmate supports a sorted check for some data types, so I figured we could rely on that when we need it.

[dajmcdon]

Do these changes result in reasonable/interpretable error messages (still get the argument names correct in the error message, etc.) Can you maybe show an example in the PR comments?

[dshemetov]

Good question right, I brought back the handle_arg_list since it handled grabbing names well. Checkmate does quite well with preserving variable names in errors.

r$> r1 <- epi_recipe(x) %>%
          step_epi_ahead(death_rate, ahead = 3.6) %>%
          step_epi_lag(death_rate, lag = 1.9)
Error in (function (name, value)  : 
  Assertion on 'ahead' failed: Must be of type 'integerish', but element 1 is not close to an integer.

r$> r2 <- epi_recipe(x) %>%
          step_epi_ahead(death_rate, ahead = 7) %>%
          step_epi_lag(death_rate, lag = -7)
Error in (function (name, value)  : 
  Assertion on 'lag' failed: Element 1 is not >= 0.

r$> step_growth_rate(r, value, role = 1)
Error in (function (name, value)  : 
  Assertion on 'role' failed: Must be of type 'character', not 'double'.

r$> step_growth_rate(r, value, horizon = 0)
Error in (function (name, value)  : 
  Assertion on 'horizon' failed: Element 1 is not >= 1.

[dajmcdon]

Perhaps for a future Issue: it would be nice if those error messages more specifically described the calling function in which the error occurred.

[dshemetov]

FWIW, traceback() can lead the user there

r$> r1 <- epi_recipe(x) %>%
              step_epi_ahead(death_rate, ahead = 3.6) %>%
              step_epi_lag(death_rate, lag = 1.9)
Error in (function (name, value)  : 
  Assertion on 'ahead' failed: Must be of type 'integerish', but element 1 is not close to an integer.

r$> traceback()
15: stop(simpleError(sprintf(msg, ...), call.))
14: mstop("Assertion on '%s' failed: %s.", var.name, res, call. = sys.call(-2L))
13: makeAssertion(x, res, .var.name, add)
12: assert_integerish(value, null.ok = allow_null, lower = 0, any.missing = FALSE, 
        .var.name = name) at utils-arg.R#63
11: (function (name, value) 
    {
        assert_integerish(value, null.ok = allow_null, lower = 0, 
            any.missing = FALSE, .var.name = name)
    })(dots[[1L]][[1L]], dots[[2L]][[1L]])
10: mapply(.f, .x, .y, MoreArgs = list(...), SIMPLIFY = FALSE) at compat-purrr.R#72
9: map2(.x, .y, .f, ...) at compat-purrr.R#15
8: walk2(names, values, .tests) at utils-arg.R#10
7: handle_arg_list(..., .tests = function(name, value) {
       assert_integerish(value, null.ok = allow_null, lower = 0, 
           any.missing = FALSE, .var.name = name)
   }) at utils-arg.R#62
6: arg_is_nonneg_int(ahead) at step_epi_shift.R#126
5: step_epi_ahead(., death_rate, ahead = 3.6)
4: inherits(x, "epi_recipe") at epi_recipe.R#221
3: is_epi_recipe(recipe) at step_epi_shift.R#64
2: step_epi_lag(., death_rate, lag = 1.9)
1: epi_recipe(x) %>% step_epi_ahead(death_rate, ahead = 3.6) %>% 
       step_epi_lag(death_rate, lag = 1.9)

[dajmcdon]

True, though it would be nice if Error in (function (name, value) : said Error in step_epi_ahead() : instead.