Fix memory_usage when calculating nested list column (#16193)

The offset column of a nested empty list column may be empty as discussed in #16164. `ListColumn.memory_usage` assumed that this column was non-empty

Unblocks rapidsai/cuspatial#1400

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - GALI PREM SAGAR (https://github.com/galipremsagar)
  - Lawrence Mitchell (https://github.com/wence-)

URL: #16193

python/cudf/cudf/core/column/lists.py

@@ -73,10 +73,15 @@ def memory_usage(self):
             child0_size = (
                 current_base_child.size + 1 - current_offset
             ) * current_base_child.base_children[0].dtype.itemsize
-            current_offset = current_base_child.base_children[
-                0
-            ].element_indexing(current_offset)
             n += child0_size
+            current_offset_col = current_base_child.base_children[0]
+            if not len(current_offset_col):
+                # See https://github.com/rapidsai/cudf/issues/16164 why
+                # offset column can be uninitialized
+                break
+            current_offset = current_offset_col.element_indexing(
+                current_offset
+            )
             current_base_child = current_base_child.base_children[1]
 
         n += (

python/cudf/cudf/tests/test_list.py

@@ -12,6 +12,7 @@
 from cudf import NA
 from cudf._lib.copying import get_element
 from cudf.api.types import is_scalar
+from cudf.core.column.column import column_empty
 from cudf.testing import assert_eq
 from cudf.testing._utils import DATETIME_TYPES, NUMERIC_TYPES, TIMEDELTA_TYPES
 
@@ -926,3 +927,29 @@ def test_list_iterate_error():
 def test_list_struct_list_memory_usage():
     df = cudf.DataFrame({"a": [[{"b": [1]}]]})
     assert df.memory_usage().sum() == 16
+
+
+def test_empty_nested_list_uninitialized_offsets_memory_usage():
+    col = column_empty(0, cudf.ListDtype(cudf.ListDtype("int64")))
+    nested_col = col.children[1]
+    empty_inner = type(nested_col)(
+        size=nested_col.size,
+        dtype=nested_col.dtype,
+        mask=nested_col.mask,
+        offset=nested_col.offset,
+        null_count=nested_col.null_count,
+        children=(
+            column_empty(0, nested_col.children[0].dtype),
+            nested_col.children[1],
+        ),
+    )
+    col_empty_offset = type(col)(
+        size=col.size,
+        dtype=col.dtype,
+        mask=col.mask,
+        offset=col.offset,
+        null_count=col.null_count,
+        children=(column_empty(0, col.children[0].dtype), empty_inner),
+    )
+    ser = cudf.Series._from_data({None: col_empty_offset})
+    assert ser.memory_usage() == 8