Ensure missing edge attributes are filled with NA instead of NULL

'plyr::rbind.fill' uses NULL to fill missing values in lists. As we now
use lists for most edge attributes, we need to handle this case
separately to ensure missing values are filled with NAs instead.

To fix this issue, we need to instantiate missing columns in dataframes
with NAs before calling 'plyr::rbind.fill'. This operation is constant
with respect to the amount of rows and should not impact performance too
much.

This works towards fixing se-sic#271.

Signed-off-by: Maximilian Löffler <[email protected]>

util-networks.R

@@ -1674,9 +1674,18 @@ merge.network.data = function(vertex.data, edge.data) {
     edge.data.filtered = Filter(function(ed) {
         return(nrow(ed) > 0)
     }, edge.data)
-    ## 2) call rbind
+    ## 2) add in missing columns
+    all.columns = Reduce(union, lapply(edge.data.filtered, colnames))
+    edge.data.filtered = lapply(edge.data.filtered, function(edges) {
+        missing.columns = setdiff(all.columns, colnames(edges))
+        for (column in missing.columns) {
+            edges[[column]] = NA
+        }
+        return(edges)
+    })
+    ## 3) call rbind
     edges = plyr::rbind.fill(edge.data.filtered)
-    ## 3) correct empty results
+    ## 4) correct empty results
     if (is.null(edges)) {
         edges = create.empty.edge.list()
     }