Fix: keep fields when unionizing empty datasets

cryosparc/dataset.py

@@ -278,7 +278,7 @@ def append_many(
         first_dset = datasets[0]
         datasets = tuple(d for d in datasets if len(d) > 0)  # skip empty datasets
         if not datasets:
-            return cls(first_dset)  # so that fields are kept
+            return cls(first_dset)  # keep the same fields as the first
 
         if not repeat_allowed:
             all_uids = n.concatenate([dset["uid"] for dset in datasets])
@@ -354,7 +354,14 @@ def union_many(
         Returns:
             Dataset: combined dataset, or empty dataset if none are provided.
         """
+        if not datasets:
+            return cls()
+
+        first_dset = datasets[0]
         datasets = tuple(d for d in datasets if len(d) > 0)  # skip empty datasets
+        if not datasets:
+            return cls(first_dset)  # keep the same fields as the first
+
         keep_fields = cls.common_fields(*datasets, assert_same_fields=assert_same_fields)
         keep_masks = []
         keep_uids = n.array([], dtype=n.uint64)

tests/test_dataset.py

@@ -296,6 +296,14 @@ def test_union_many_empty():
     assert len(Dataset.union_many().rows()) == 0
 
 
+def test_union_empty_fields():
+    d1 = Dataset.allocate(0, [("field", "f4")])
+    d2 = Dataset.allocate(0, [("field", "f4")])
+    d3 = d1.union(d2)
+    assert len(d3) == 0
+    assert d3.fields() == ["uid", "field"]
+
+
 def test_allocate_many_separate():
     for _ in range(66_000):
         allocated = []