fix: preserve falsy but valid multi-index level names

doc/whats-new.rst

@@ -30,6 +30,12 @@ Bug Fixes
   differ from the dtype returned by the matching numpy-backed bottleneck path,
   notably ``object`` instead of ``float64`` for boolean inputs.
   By `Matthew Rocklin <https://github.com/mrocklin>`_.
+- Fixed a :py:class:`KeyError` when building a multi-index (e.g. via
+  :py:meth:`Dataset.stack` or :py:meth:`Dataset.set_index`) from levels whose
+  name is falsy but valid, such as an empty string ``""``. Such names were
+  incorrectly replaced by the default ``{dim}_level_{i}`` name instead of being
+  preserved.
+  By `Kropiunig <https://github.com/Kropiunig>`_.
 
 
 Documentation

xarray/core/indexes.py

@@ -1038,7 +1038,7 @@ def __init__(self, array: Any, dim: Hashable, level_coords_dtype: Any = None):
         # default index level names
         names = []
         for i, idx in enumerate(self.index.levels):
-            name = idx.name or f"{dim}_level_{i}"
+            name = idx.name if idx.name is not None else f"{dim}_level_{i}"
             if name == dim:
                 raise ValueError(
                     f"conflicting multi-index level name {name!r} with dimension {dim!r}"

xarray/tests/test_indexes.py

@@ -354,6 +354,12 @@ def test_constructor(self) -> None:
         index = PandasMultiIndex(pd_idx, "x")
         assert list(index.index.names) == ["x_level_0", "x_level_1"]
 
+        # a falsy but valid level name (e.g. "") must be preserved rather than
+        # replaced by the default ``{dim}_level_{i}`` name
+        pd_idx = pd.MultiIndex.from_arrays([foo_data, bar_data], names=("", "bar"))
+        index = PandasMultiIndex(pd_idx, "x")
+        assert list(index.index.names) == ["", "bar"]
+
     def test_from_variables(self) -> None:
         v_level1 = xr.Variable(
             "x", [1, 2, 3], attrs={"unit": "m"}, encoding={"dtype": np.int32}