fix

packages/vaex-core/vaex/dataset.py

@@ -753,6 +753,9 @@ def _decode(cls, encoding, spec):
         return ds
 
     def is_masked(self, column):
+        for dataset in self.datasets:
+            if column not in dataset:
+                return True
         return any(k.is_masked(column) for k in self.datasets)
 
     def shape(self, column):

tests/concat_test.py

@@ -238,6 +238,8 @@ def test_concat_unaligned_schema(arrow):
     df = df1.concat(df2)
     assert df.x.tolist() == [1, 2, None, None]
     assert df.y.tolist() == [None, None, 'd', 'e']
+    assert df.is_masked('x')
+    assert df.is_masked('y')
     # always 'upcast' to Arrow arrays
     # # rationale: Arrow will use use less memory, numpy has no efficient way to represent all missing data
     assert df.x.data_type() == pa.float32()

tests/export_test.py

@@ -117,7 +117,7 @@ def test_export_concat_missing_cols_hdf5(tmpdir):
     df.export_hdf5(filename)
 
     df_opened = vaex.open(filename)
-    assert df_opened.x.tolist() == [1, 2, 3, 7, 8, 9]
+    assert df_opened.x.tolist() == [1, 2, 3, 4, 5, 6]
     assert df_opened.y.tolist() == [10, 20, 30, None, None, None]
     assert df_opened.s.tolist() == ['x1', 'x2', 'x3', None, None, None]