Avoid segfault when fetching sparse vectors with flexible dimensions.

Author: anthony-tuininga

doc/src/user_guide/dataframes.rst

@@ -174,7 +174,8 @@ When converting Oracle Database VECTORs:
   ``indices`` and ``values`` similar to :ref:`SparseVector objects
   <sparsevectorsobj>`.
 
-- VECTOR columns with flexible dimensions are supported.
+- Fixed and flexible dimensions are supported for dense VECTOR columns. For
+  sparse VECTOR columns, the dimension of each vector must be the same.
 
 - VECTOR columns with flexible formats are not supported. Each vector value
   must have the same storage format data type.

src/oracledb/errors.py

@@ -288,6 +288,7 @@ def _raise_not_supported(feature: str) -> None:
 ERR_PAYLOAD_CANNOT_BE_ENQUEUED = 2062
 ERR_SCROLL_OUT_OF_RESULT_SET = 2063
 ERR_POOL_MAX_LESS_THAN_MIN = 2064
+ERR_ARROW_SPARSE_VECTOR_NOT_ALLOWED = 2065
 
 # error numbers that result in NotSupportedError
 ERR_TIME_NOT_SUPPORTED = 3000
@@ -893,6 +894,10 @@ def _raise_not_supported(feature: str) -> None:
     ERR_INVALID_NETWORK_NAME: (
         '"{name}" includes characters that are not allowed'
     ),
+    ERR_ARROW_SPARSE_VECTOR_NOT_ALLOWED: (
+        "Apache Arrow format does not support sparse vectors with flexible "
+        "dimensions"
+    ),
     ERR_ARROW_UNSUPPORTED_DATA_TYPE: (
         "conversion from Oracle Database type {db_type_name} to Apache "
         "Arrow format is not supported"

src/oracledb/interchange/nanoarrow_bridge.pyx

@@ -604,26 +604,30 @@ cdef class OracleArrowArray:
         """
         Append a sparse vector to the array.
         """
-        cdef:
-            ArrowArray *num_dims_array = self.arrow_array.children[0]
-            ArrowArray *indices_array = self.arrow_array.children[1]
-            ArrowArray *values_array = self.arrow_array.children[2]
+        cdef ArrowArray *array
+
+        # validate that the array supports sparse vectors
+        if self.arrow_type != NANOARROW_TYPE_STRUCT:
+            errors._raise_err(errors.ERR_ARROW_SPARSE_VECTOR_NOT_ALLOWED)
 
         # append number of dimensions
-        _check_nanoarrow(ArrowArrayAppendInt(num_dims_array, num_dims))
+        array = self.arrow_array.children[0]
+        _check_nanoarrow(ArrowArrayAppendInt(array, num_dims))
 
         # append indices array
-        append_uint32_array(indices_array, indices)
+        array = self.arrow_array.children[1]
+        append_uint32_array(array, indices)
 
         # append values array
+        array = self.arrow_array.children[2]
         if self.child_arrow_type == NANOARROW_TYPE_FLOAT:
-            append_float_array(values_array, values)
+            append_float_array(array, values)
         elif self.child_arrow_type == NANOARROW_TYPE_DOUBLE:
-            append_double_array(values_array, values)
+            append_double_array(array, values)
         elif self.child_arrow_type == NANOARROW_TYPE_INT8:
-            append_int8_array(values_array, values)
+            append_int8_array(array, values)
         elif self.child_arrow_type == NANOARROW_TYPE_UINT8:
-            append_uint8_array(values_array, values)
+            append_uint8_array(array, values)
 
         # indicate structure is completed
         _check_nanoarrow(ArrowArrayFinishElement(self.arrow_array))

tests/test_8000_dataframe.py

@@ -1119,6 +1119,34 @@ def test_8038(self):
                 """
             )
 
+    @unittest.skipUnless(
+        test_env.has_client_version(23, 7), "unsupported client"
+    )
+    @unittest.skipUnless(
+        test_env.has_server_version(23, 7), "unsupported server"
+    )
+    def test_8039(self):
+        "8039 - DPY-4007 -fetch sparse vectors with flexible dimensions"
+        self.__check_interop()
+        with self.assertRaisesFullCode("DPY-2065"):
+            self.conn.fetch_df_all(
+                """
+                SELECT TO_VECTOR(
+                    TO_VECTOR('[34.6, 0, 0, 0, 0, 0, 77.8]', 7, FLOAT64),
+                    7,
+                    FLOAT64,
+                    SPARSE
+                    )
+                union all
+                SELECT TO_VECTOR(
+                    TO_VECTOR('[34.6, 0, 0, 0, 0, 0, 0, 9.1]', 8, FLOAT64),
+                    8,
+                    FLOAT64,
+                    SPARSE
+                    )
+                """
+            )
+
 
 if __name__ == "__main__":
     test_env.run_test_cases()