handle duplicates in insertion into trie properly (return same trie if it already has a FOUND/FOUND_WITH_CHILDREN node awith that key, or change tag of node from CHILDREN to FOUND_WITH_CHILDREN to mark it as inserted

src/string-set.sml

@@ -416,9 +416,24 @@ struct
         * the tag/case. *)
        | FULL_INSERT_MATCH =>
            (* in case of a full match, 
-            * returned node should always be FOUND_WITH_CHILDREN,
+            * need to turn child into FOUND_WITH_CHILDREN
-            * because full match means this key was inserted into the trie. *)
+            * or FOUND case, if it not already either
-           FOUND_WITH_CHILDREN {keys = keys, children = children}
+            *)
+           (case Vector.sub (children, insIdx) of
+              CHILDREN {keys = childKeys, children = childChildren} =>
+                let
+                  val newTrieChild =
+                    FOUND_WITH_CHILDREN
+                      {keys = childKeys, children = childChildren}
+                  val newParentChildren =
+                    Vector.mapi
+                      (fn (childIdx, elt) =>
+                         if insIdx <> childIdx then elt else newTrieChild)
+                      children
+                in
+                  constructor {keys = keys, children = newParentChildren}
+                end
+            | _ => trie)
        (* if insert key contains trie key, need to recurse down node *)
        | INSERT_KEY_CONTAINS_TRIE_KEY =>
            let
@@ -572,18 +587,16 @@ struct
         in
           CHANGED newNode
         end
-    else 
+    else (* if the caller was from the FOUND_WITH_CHILDREN case,
-      (* if the caller was from the FOUND_WITH_CHILDREN case,
+          * then, instead of deleting this node entirely, 
-       * then, instead of deleting this node entirely, 
+          * we only delete this node's children, and mark this node 
-       * we only delete this node's children, and mark this node 
+          * as found. 
-       * as found. 
+          * However, in general case where this is a CHILDREN node 
-       * However, in general case where this is a CHILDREN node 
+          * whose key was not inserted into the trie, 
-       * whose key was not inserted into the trie, 
+          * we should fully delete this node as well. *) if isFoundWithChildren then
-       * we should fully delete this node as well. *)
+      CHANGED FOUND
-      if isFoundWithChildren then
+    else
-        CHANGED FOUND
+      MADE_EMPTY
-      else
-        MADE_EMPTY
 
   (* should be called when searchKeyMatch returns FULL_MATCH
    *in helpRemove function *)

tests/string-set-tests.sml

@@ -136,6 +136,17 @@ struct
       print "StringSet.remove: passed remove5\n"
     end
 
+  fun insert1 () = 
+    let
+      val trie = StringSet.empty
+      val _ = assertFalse (StringSet.exists ("abc", trie), "abc does not exist before it is added to trie")
+
+      val trie = StringSet.insert ("abc", trie)
+      val _ = assertTrue (StringSet.exists ("abc", trie), "abc exists after being addedd to trie")
+    in
+      print "StringSet.insert: passed add1\n"
+    end
+
   fun run () =
     let
       val _ = testExists ()
@@ -146,6 +157,8 @@ struct
       val _ = remove3 ()
       val _ = remove4 ()
       val _ = remove5 ()
+
+      val _ = insert1 ()
     in
       ()
     end