Profile the time issue for arx_epi_forecaster() on an epi_archive

[rachlobay]

arx_epi_forecaster() is noticeably slower than the old arx_forecaster() on an epi_archive object (EDIT: I see the forecasters are now both called arx_forecaster(), but to make it clear which I'm referring to, I'll use the old names here). So, the time issue for `arx_epi_forecaster() should be examined... Ex. What's causing the forecaster to be slower (and what can we do about that)?

An example is given below to compare arx_epi_forecaster() to arx_forecaster() on an epi_archive object. Note that this example can be used and profiling can proceed after the tibble::as_tibble() %>% line is removed in the slide function for an archive in epiprocess (see Issue #208). We probably don't have to wait until that issue is fixed to start profiling, but the person who addresses this issue should then use a branch of epiprocess where the problem line in slide is removed (else if you run the below ex, you'll get an error: Error: epi_data must be an epi_df).

library(epipredict)
library(epiprocess)
library(covidcast)
library(data.table)
library(dplyr)
library(tidyr)
library(ggplot2)

y <- covidcast_signals(
  c("doctor-visits", "jhu-csse"),
  c("smoothed_adj_cli", "confirmed_7dav_incidence_prop"),
  start_day = "2020-06-01",
  end_day = "2021-12-01",
  issues = c("2020-06-01", "2021-12-01"),
  geo_type = "state",
  geo_values = c("ca", "fl"))

z <- y[[1]] %>%
  select(geo_value, time_value, version = issue, percent_cli = value) %>%
  as_epi_archive()

z <- epix_merge(
  z, y[[2]] %>%
    select(geo_value, time_value, version = issue, case_rate = value) %>%
    as_epi_archive(), sync = "locf")

fc_time_values <- seq(as.Date("2020-08-01"), as.Date("2021-12-01"),
                      by = "1 month")
ahead = 7

# Old arx_forecaster is pretty quick
z %>% epix_slide(fc = arx_forecaster(y = case_rate, 
                                     key_vars = geo_value, time_value = time_value,
                                     args = arx_args_list(ahead = ahead)),
                 n = 120, ref_time_values = fc_time_values)

# New arx_epi_forecaster is noticeably slower?
z %>%
  epix_slide(function(x, ...)
    arx_epi_forecaster(x, outcome = "case_rate",
                       predictors = c("case_rate"),
                       args = arx_args_list(ahead = ahead))$predictions %>%
      select(-c(geo_value, time_value)),
    n = 120, ref_time_values = fc_time_values, new_col_name = "fc")