New test added to reveal a bug involving encoding splits on pruned tr…

…ees.

If the first taxon is not in the tree when the splits are encoded,
the mask in NormalizedBitmaskDict will be wrong. Failure to have the
1-bit set will cause the mask to be complemented. Subsequent "anding"
of that mask to any split's bitmask will result in a 0 key in the
NormalizedBitmaskDict. Thus there will be only one key in the dict and
bitmask-dependent operations (such as clade summarization and tree-to-tree
distances) will be incorrect.

dendropy/test/test_splits_on_incomplete_leaf_sets.py

@@ -57,6 +57,25 @@ def testUnrooted(self):
     def testRooted(self):
         self.check("Rooted", "incomplete_leaves_rooted")
 
+    def testPrunedThenEncoding(self):
+        from cStringIO import StringIO
+        FIX = False
+        inp = StringIO('''(a,b,c,(d,e));
+        (b,d,(c,e));''')
+        first, second = dendropy.TreeList.get_from_stream(inp, schema='newick')
+        # prune tree 1 to have the same leaf set as tree 2.
+        #   this removes the first taxon in the taxon list "A"
+        retain_list = set([node.taxon for node in second.leaf_nodes()])
+        exclude_list = [node for node in first.leaf_nodes() if node.taxon not in retain_list]
+        for nd in exclude_list:
+            first.prune_subtree(nd)
+        # the trees are now (b,c,(d,e)) and (b,d,(c,e)) so the symmetric diff is 2
+        if FIX:
+            dendropy.treesplit.encode_splits(first, lowest_relevant_bit=2)
+            dendropy.treesplit.encode_splits(second, lowest_relevant_bit=2)
+        self.assertEquals(2, first.symmetric_difference(second))
+
+
 if __name__ == "__main__":
     unittest.main()