Taking Random Subsets of Dataframes in R

R is very inefficient (use Julia or your algorithms are or it’s sometimes useful anyway to run everything with random subsets of the whole dataset. tidyverse didn’t seem to have a straight-forward function so I wrote this.

# It's often more reliable to hang onto the ID variables even if you need to always
# handle them separately within the function. You don't want to suddenly mix IDs.
subset_random <- function(data, id_var_names, max_variables, max_observations) {
 
  if (ncol(data) > max_variables) {
    
    n_id_vars <- length(id_var_names)
    n_cols <- ncol(data)
    
    data <- data %>%
      # Sample only from non-ID variables.
      select( 
	# ID variables first if not already like that.
        all_of(id_var_names), 
        everything()
      ) %>%
      select(                           
	# A sample of indeces after ID variables.
        base::sample(x = seq(n_id_vars + 1, n_cols), size = max_variables)
      )
  }
  
  if (nrow(data) > max_observations) {                                                   
    data <- data %>% 
      slice_sample(n = max_observations)
    
  }
 
  return(data)
}