Optimize calculation of combination names

Makes calling extractCombinationsImpl about 3x faster

R/data-helpers.R

@@ -190,19 +190,23 @@ extractCombinationsImpl <- function(df,
   cc <- colorLookup(colors)
 
   elems <- rownames(df)
-  dfSets <- df[, allSetNames]
-  cName <- apply(dfSets, 1, function(r) {
-    nn <- allSetNames[as.logical(r)]
-    if (length(nn) == 1) {
-      nn
-    } else {
-      paste(nn, collapse = symbol)
-    }
+
+  # Calculate combinations names.
+  # First, translate from 1/0 per set for member or not into the name of each
+  # set + separating symbol if member, or empty string if not.
+  translatedSets <- lapply(allSetNames, function(setName) {
+    # Same as ifelse(df[[setName]] == 1, paste0(setName, symbol), ""), but a lot faster
+    c("", paste0(setName, symbol))[df[[setName]] + 1]
   })
+  # Then paste0() these translated names by row
+  cName <- do.call(paste0, translatedSets)
+
   dd <- aggregate(elems, list(c_name = cName), function(r) {
     r
   })
   setNames <- strsplit(dd$c_name, symbol, fixed = TRUE)
+  # We got an extra symbol with the translatedSets above; clean it up
+  dd$c_name <- vapply(setNames, paste0, collapse = symbol, character(1))
   setColors <- cc(dd$c_name)
 
   combinations <- lapply(seq_len(nrow(dd)), function(i) {