(Gaps causing) missing test predictors gives confusing error

[brookslogan]

library(epipredict)
#> Loading required package: epiprocess
#> Registered S3 method overwritten by 'tsibble':
#>   method               from 
#>   as_tibble.grouped_df dplyr
#> 
#> Attaching package: 'epiprocess'
#> The following object is masked from 'package:stats':
#> 
#>     filter
#> Loading required package: parsnip
#> Registered S3 method overwritten by 'epipredict':
#>   method            from   
#>   print.step_naomit recipes
jhu <- case_death_rate_subset %>%
  dplyr::filter(time_value >= as.Date("2021-12-01")) %>%
  dplyr::filter(time_value != max(time_value) - 7L)

out <- arx_forecaster(
  jhu, "death_rate",
  c("case_rate", "death_rate")
)
#> Error in `as_list_of()` at epipredict/R/dist_quantiles.R:63:3:
#> ! Can't find common type for elements of `x`.

^{Created on 2024-07-17 with reprex v2.1.1}

Just skimming, it looks like there might be one underlying issue + a similar related issue in a separate branch of the same function:

• get_test_data() with fill_locf = TRUE does not complete() before fill()-ing. To avoid this error, we'd might want it to complete() according to the requested lags and/or a guessed period(... or maybe restructure prediction entirely to somehow fill() and select test instances after baking the test data?) --- this appears to be the case we are in with arx_forecaster()
• get_test_data() with fill_locf = FALSE probably doesn't raise a comprehensible error message in this case either. --- this is a related error case

[brookslogan]

Potential duplicate of #333? [Different error message though.]

[brookslogan]

Once I fixed the example in #333, this does appear to be a duplicate. Closing.