Performance boost?

src/lib.rs

@@ -253,10 +253,14 @@ impl<'a> ops::Sub<&'a BitSet> for &'a BitSet {
     fn sub(self, rhs: &'a BitSet) -> BitSet {
         let mut set = self.clone();
 
-        for (key, value) in &rhs.blocks {
-            let block = set.blocks.entry(*key).or_insert(0);
+        for (key, block) in &mut set.blocks {
+            if let Some(value) = rhs.blocks.get(&key) {
+                // let n = block.count_ones();
 
                 *block &= !value;
+
+                // set.nbits += (block.count_ones() - n) as usize;
+            }
         }
 
         set.nbits = set.blocks.values().map(|block| block.count_ones()).sum::<u32>() as usize;