Caching does not work for `xy_dgp_constructor`

[tiffanymtang]

Caching does not work when using xy_dgp_constructor() to construct a DGP. When comparing the cached experiment parameters to the current experiment parameters (i.e., in n_reps_cached <- private$.n_reps_cached(cached_params$fit)), the result is that none of the current experiment parameters have been previously cached and so n_reps_cached = 0. I think the .dgp_params column in the tibble is the issue. Parameters like x_fun and y_fun cannot be easily compared because they are functions with their own environments.

Example:

dgp <- create_dgp(
  .dgp_fun = xy_dgp_constructor,
  .name = 'Uncorrelated Gaussian Linear DGP',
  x_fun = generate_X_gaussian,
  y_fun = generate_y_linear,
  n = 200,
  p = 10,
  betas = c(rep(1, 5), rep(0, 5))
)
method <- create_method(
  .method_fun = function(X, y, support) return(rnorm(1))
)

experiment <- create_experiment() %>%
  add_dgp(dgp) %>%
  add_method(method)

results <- run_experiment(experiment, save = TRUE)
cached_results <- run_experiment(experiment, use_cached = TRUE, save = FALSE)
all.equal(results, cached_results)

[jpdunc23]

Moved to dgpoix.