perf: Only calculate data group interactions once

This goes towards making overall grouped calculations quicker. See #119
nathaneastwood · Dec 29, 2022 · 3cc0a99 · 3cc0a99
1 parent e910fd7
commit 3cc0a99
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/R/group_utils.R b/R/group_utils.R
@@ -64,8 +64,9 @@ calculate_groups <- function(data, groups, drop = group_by_drop_default(data)) {
   is_factor <- do.call(c, lapply(unique_groups, function(x) is.factor(x)))
   n_comb <- nrow(unique_groups)
   rows <- rep(list(NA), n_comb)
+  data_groups <- interaction(data[, groups, drop = TRUE])
   for (i in seq_len(n_comb)) {
-    rows[[i]] <- which(interaction(data[, groups, drop = TRUE]) %in% interaction(unique_groups[i, groups]))
+    rows[[i]] <- which(data_groups %in% interaction(unique_groups[i, groups]))
   }
 
   if (!isTRUE(drop) && any(is_factor)) {