fix: Fix too many cluster columns requested by caching (#2155)

Author: TrevorBergeron

bigframes/session/bq_caching_executor.py

@@ -637,14 +637,18 @@ def _execute_plan_gbq(
 
             create_table = True
             if not cache_spec.cluster_cols:
-                assert len(cache_spec.cluster_cols) <= _MAX_CLUSTER_COLUMNS
                 offsets_id = bigframes.core.identifiers.ColumnId(
                     bigframes.core.guid.generate_guid()
                 )
                 plan = nodes.PromoteOffsetsNode(plan, offsets_id)
                 cluster_cols = [offsets_id.sql]
             else:
-                cluster_cols = cache_spec.cluster_cols
+                cluster_cols = [
+                    col
+                    for col in cache_spec.cluster_cols
+                    if bigframes.dtypes.is_clusterable(plan.schema.get_type(col))
+                ]
+                cluster_cols = cluster_cols[:_MAX_CLUSTER_COLUMNS]
 
         compiled = compile.compile_sql(
             compile.CompileRequest(

tests/system/small/test_dataframe.py

@@ -5537,6 +5537,23 @@ def test_df_cached(scalars_df_index):
     pandas.testing.assert_frame_equal(df.to_pandas(), df_cached_copy.to_pandas())
 
 
+def test_df_cached_many_index_cols(scalars_df_index):
+    index_cols = [
+        "int64_too",
+        "time_col",
+        "int64_col",
+        "bool_col",
+        "date_col",
+        "timestamp_col",
+        "string_col",
+    ]
+    df = scalars_df_index.set_index(index_cols)
+    df = df[df["rowindex_2"] % 2 == 0]
+
+    df_cached_copy = df.cache()
+    pandas.testing.assert_frame_equal(df.to_pandas(), df_cached_copy.to_pandas())
+
+
 def test_assign_after_binop_row_joins():
     pd_df = pd.DataFrame(
         {