diff --git a/src/gj.rs b/src/gj.rs
index e6e5e0ae..a2150387 100644
--- a/src/gj.rs
+++ b/src/gj.rs
@@ -719,32 +719,13 @@ impl EGraph {
                 }
             }
 
-            let is_rebuilding = cq.query.ruleset.to_string().contains("rebuilding_");
             let do_seminaive = self.seminaive && !global_updated;
             // for the later atoms, we consider everything
             let mut timestamp_ranges = vec![0..u32::MAX; cq.query.atoms.len()];
             if do_seminaive {
-                for (atom_i, atom) in cq.query.atoms.iter().enumerate() {
+                for (atom_i, _atom) in cq.query.atoms.iter().enumerate() {
                     timestamp_ranges[atom_i] = timestamp..u32::MAX;
 
-                    // For rebuilding, we have the invariant
-                    // that new atoms are up-to-date w.r.t.
-                    // old unionfind entries.
-                    // So do the join for new unionfind entries
-                    // and all the other atoms.
-                    // TODO this hack fails on one benchmark, not sure why
-                    let rebuilding_hack = false;
-                    if rebuilding_hack {
-                        let atom_has_parent = format!("{:?}", atom).contains("Parent_");
-                        if is_rebuilding && !atom_has_parent {
-                            continue;
-                        } else if is_rebuilding {
-                            assert_eq!(cq.query.atoms.len(), 2);
-                            timestamp_ranges = vec![0..u32::MAX; cq.query.atoms.len()];
-                            timestamp_ranges[atom_i] = timestamp..u32::MAX;
-                        }
-                    }
-
                     self.gj_for_atom(Some(atom_i), &timestamp_ranges, cq, &mut f);
                     // now we can fix this atom to be "old stuff" only
                     // range is half-open; timestamp is excluded
diff --git a/src/lib.rs b/src/lib.rs
index b311fbb3..57c29f52 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -946,7 +946,7 @@ impl EGraph {
         let name = Symbol::from(name);
         let mut ctx = typecheck::Context::new(self);
         let (query0, action0) = ctx
-            .typecheck_query(&rule.body, &rule.head, ruleset)
+            .typecheck_query(&rule.body, &rule.head)
             .map_err(Error::TypeErrors)?;
         let query = self.compile_gj_query(query0, &ctx.types);
         let program = self
@@ -1048,7 +1048,7 @@ impl EGraph {
         let converted_facts = facts.iter().map(|f| f.to_fact()).collect::<Vec<Fact>>();
         let empty_actions = vec![];
         let (query0, _) = ctx
-            .typecheck_query(&converted_facts, &empty_actions, Symbol::from(""))
+            .typecheck_query(&converted_facts, &empty_actions)
             .map_err(Error::TypeErrors)?;
         let query = self.compile_gj_query(query0, &ctx.types);
 
@@ -1221,7 +1221,7 @@ impl EGraph {
         let function_type = self
             .type_info()
             .lookup_user_func(func_name)
-            .unwrap_or_else(|| panic!("Unrecognzed function name {}", func_name));
+            .unwrap_or_else(|| panic!("Unrecognized function name {}", func_name));
         let func = self.functions.get_mut(&func_name).unwrap();
 
         let mut filename = self.fact_directory.clone().unwrap_or_default();
diff --git a/src/sort/map.rs b/src/sort/map.rs
index d270a703..c8abacb6 100644
--- a/src/sort/map.rs
+++ b/src/sort/map.rs
@@ -87,8 +87,22 @@ impl Sort for MapSort {
         result
     }
 
-    fn canonicalize(&self, _value: &mut Value, _unionfind: &UnionFind) -> bool {
-        false
+    fn canonicalize(&self, value: &mut Value, unionfind: &UnionFind) -> bool {
+        let maps = self.maps.lock().unwrap();
+        let map = maps.get_index(value.bits as usize).unwrap();
+        let mut changed = false;
+        let new_map: ValueMap = map
+            .iter()
+            .map(|(k, v)| {
+                let (mut k, mut v) = (*k, *v);
+                changed |= self.key.canonicalize(&mut k, unionfind);
+                changed |= self.value.canonicalize(&mut v, unionfind);
+                (k, v)
+            })
+            .collect();
+        drop(maps);
+        *value = new_map.store(self).unwrap();
+        changed
     }
 
     fn register_primitives(self: Arc<Self>, typeinfo: &mut TypeInfo) {
@@ -194,16 +208,9 @@ impl PrimitiveLike for MapRebuild {
     fn apply(&self, values: &[Value], egraph: &EGraph) -> Option<Value> {
         let maps = self.map.maps.lock().unwrap();
         let map = maps.get_index(values[0].bits as usize).unwrap();
-        let mut changed = false;
         let new_map: ValueMap = map
             .iter()
-            .map(|(k, v)| {
-                let (k, v) = (*k, *v);
-                let updated_k = egraph.find(k);
-                let updated_v = egraph.find(v);
-                changed |= updated_k != k || updated_v != v;
-                (updated_k, updated_v)
-            })
+            .map(|(k, v)| (egraph.find(*k), egraph.find(*v)))
             .collect();
 
         drop(maps);
diff --git a/src/sort/set.rs b/src/sort/set.rs
index c6a9b437..b25d7fe4 100644
--- a/src/sort/set.rs
+++ b/src/sort/set.rs
@@ -86,8 +86,21 @@ impl Sort for SetSort {
         result
     }
 
-    fn canonicalize(&self, _value: &mut Value, _unionfind: &UnionFind) -> bool {
-        false
+    fn canonicalize(&self, value: &mut Value, unionfind: &UnionFind) -> bool {
+        let sets = self.sets.lock().unwrap();
+        let set = sets.get_index(value.bits as usize).unwrap();
+        let mut changed = false;
+        let new_set: ValueSet = set
+            .iter()
+            .map(|e| {
+                let mut e = *e;
+                changed |= self.element.canonicalize(&mut e, unionfind);
+                e
+            })
+            .collect();
+        drop(sets);
+        *value = new_set.store(self).unwrap();
+        changed
     }
 
     fn register_primitives(self: Arc<Self>, typeinfo: &mut TypeInfo) {
diff --git a/src/sort/vec.rs b/src/sort/vec.rs
index 1565f6b0..be97ae7d 100644
--- a/src/sort/vec.rs
+++ b/src/sort/vec.rs
@@ -87,7 +87,7 @@ impl Sort for VecSort {
         let vecs = self.vecs.lock().unwrap();
         let vec = vecs.get_index(value.bits as usize).unwrap();
         let mut changed = false;
-        let new_set: ValueVec = vec
+        let new_vec: ValueVec = vec
             .iter()
             .map(|e| {
                 let mut e = *e;
@@ -96,7 +96,7 @@ impl Sort for VecSort {
             })
             .collect();
         drop(vecs);
-        *value = new_set.store(self).unwrap();
+        *value = new_vec.store(self).unwrap();
         changed
     }
 
@@ -215,17 +215,9 @@ impl PrimitiveLike for VecRebuild {
     fn apply(&self, values: &[Value], egraph: &EGraph) -> Option<Value> {
         let vec = ValueVec::load(&self.vec, &values[0]);
 
-        let mut changed = false;
-        let new_set: ValueVec = vec
-            .iter()
-            .map(|e| {
-                let updated = egraph.find(*e);
-                changed |= updated != *e;
-                updated
-            })
-            .collect();
+        let new_vec: ValueVec = vec.iter().map(|e| egraph.find(*e)).collect();
         drop(vec);
-        Some(new_set.store(&self.vec).unwrap())
+        Some(new_vec.store(&self.vec).unwrap())
     }
 }
 struct VecOf {
diff --git a/src/typecheck.rs b/src/typecheck.rs
index 1b9012f4..d0606c37 100644
--- a/src/typecheck.rs
+++ b/src/typecheck.rs
@@ -52,8 +52,6 @@ impl<T: std::fmt::Display> std::fmt::Display for Atom<T> {
 pub struct Query {
     pub atoms: Vec<Atom<Symbol>>,
     pub filters: Vec<Atom<Primitive>>,
-    // store the ruleset for hack on rebuilding
-    pub(crate) ruleset: Symbol,
 }
 
 impl std::fmt::Display for Query {
@@ -131,7 +129,6 @@ impl<'a> Context<'a> {
         &mut self,
         facts: &'a [Fact],
         actions: &'a [Action],
-        ruleset: Symbol,
     ) -> Result<(Query, Vec<Action>), Vec<TypeError>> {
         for fact in facts {
             self.typecheck_fact(fact);
@@ -215,7 +212,6 @@ impl<'a> Context<'a> {
         let mut query = Query {
             atoms: Default::default(),
             filters: Default::default(),
-            ruleset,
         };
         let mut query_eclasses = HashSet::<Id>::default();
         // Now we can fill in the nodes with the canonical leaves
@@ -729,11 +725,15 @@ impl EGraph {
     fn perform_set(
         &mut self,
         table: Symbol,
-        args: &[Value],
         new_value: Value,
         stack: &mut Vec<Value>,
     ) -> Result<(), Error> {
         let function = self.functions.get_mut(&table).unwrap();
+
+        let new_len = stack.len() - function.schema.input.len();
+        // TODO would be nice to use slice here
+        let args = &stack[new_len..];
+
         // We should only have canonical values here: omit the canonicalization step
         let old_value = function.get(args);
 
@@ -757,6 +757,7 @@ impl EGraph {
                     }
                 };
                 if merged != old_value {
+                    let args = &stack[new_len..];
                     let function = self.functions.get_mut(&table).unwrap();
                     function.insert(args, merged, self.timestamp);
                 }
@@ -864,10 +865,8 @@ impl EGraph {
                     // except for setting the parent relation
                     let new_value = stack.pop().unwrap();
                     let new_len = stack.len() - function.schema.input.len();
-                    // TODO would be nice to use slice here
-                    let args = stack[new_len..].to_vec();
 
-                    self.perform_set(*f, &args, new_value, stack)?;
+                    self.perform_set(*f, new_value, stack)?;
                     stack.truncate(new_len)
                 }
                 Instruction::Union(arity) => {
diff --git a/src/typechecking.rs b/src/typechecking.rs
index d5057322..8aa2090b 100644
--- a/src/typechecking.rs
+++ b/src/typechecking.rs
@@ -20,7 +20,7 @@ pub struct TypeInfo {
     pub presort_names: HashSet<Symbol>,
     pub sorts: HashMap<Symbol, Arc<dyn Sort>>,
     pub primitives: HashMap<Symbol, Vec<Primitive>>,
-    func_types: HashMap<Symbol, FuncType>,
+    pub func_types: HashMap<Symbol, FuncType>,
     global_types: HashMap<Symbol, ArcSort>,
     pub local_types: HashMap<CommandId, HashMap<Symbol, ArcSort>>,
 }
diff --git a/tests/levenshtein-distance.egg b/tests/levenshtein-distance.egg
index 59f234e0..c80862cd 100644
--- a/tests/levenshtein-distance.egg
+++ b/tests/levenshtein-distance.egg
@@ -60,5 +60,11 @@
 
 (run 100)
 
+(extract (Unwrap Test1))
+(check (= Test1 (Num 3)))
 
-(extract (Unwrap Test3))
\ No newline at end of file
+(extract (Unwrap Test2))
+(check (= Test2 (Num 5)))
+
+(extract (Unwrap Test3))
+(check (= Test3 (Num 5)))
\ No newline at end of file