Avoid loading entire dataset by getting the nbytes in an array

doc/whats-new.rst

@@ -35,6 +35,9 @@ Deprecations
 Bug fixes
 ~~~~~~~~~
 
+- Accessing the property ``.nbytes`` of a DataArray, or Variable no longer
+  accidentally triggers loading the variable into memory.
+
 
 Documentation
 ~~~~~~~~~~~~~

xarray/core/variable.py

@@ -403,8 +403,8 @@ def nbytes(self) -> int:
         If the underlying data array does not include ``nbytes``, estimates
         the bytes consumed based on the ``size`` and ``dtype``.
         """
-        if hasattr(self.data, "nbytes"):
-            return self.data.nbytes
+        if hasattr(self._data, "nbytes"):
+            return self._data.nbytes
         else:
             return self.size * self.dtype.itemsize
 

xarray/tests/test_dataarray.py

@@ -30,6 +30,7 @@
 from xarray.core.types import QueryEngineOptions, QueryParserOptions
 from xarray.core.utils import is_scalar
 from xarray.tests import (
+    InaccessibleArray,
     ReturnItem,
     assert_allclose,
     assert_array_equal,
@@ -3277,6 +3278,21 @@ def test_from_multiindex_series_sparse(self) -> None:
 
         np.testing.assert_equal(actual_coords, expected_coords)
 
+    def test_nbytes_does_not_load_data(self) -> None:
+        array = InaccessibleArray(np.zeros((3, 3), dtype="uint8"))
+        da = xr.DataArray(array, dims=["x", "y"])
+
+        # If xarray tries to instantiate the InaccessibleArray to compute
+        # nbytes, the following will raise an error.
+        # However, it should still be able to accurately give us information
+        # about the number of bytes from the metadata
+        assert da.nbytes == 9
+        # Here we confirm that this does not depend on array having the
+        # nbytes property, since it isn't really required by the array
+        # interface. nbytes is more a property of arrays that have been
+        # cast to numpy arrays.
+        assert not hasattr(array, "nbytes")
+
     def test_to_and_from_empty_series(self) -> None:
         # GH697
         expected = pd.Series([], dtype=np.float64)