Use User Data Subsets for faster processing in Drake Workflow

user.R

drake_plan(
  # dplyr::distinct instead?
  user_data = tidy_data %>% unique(user_id) %>% select(user_related_cols), # remove "tweet level" vars for faster runtime
  
  geo_data = do_user_data_stuff(),
  
  
  # left_join: Join matching rows from b to a.
  joined_data = target(command = left_join(user_data, tidy_data, by = "user_id"),
                       format = "fst" # useful ???
                       )

)