Ensure that a helpful error message is raised when using Apache Arrow

fixed width binary with data that doesn't conform to the fixed width.

Author: anthony-tuininga

src/oracledb/errors.py

@@ -372,6 +372,7 @@ def _raise_not_supported(feature: str) -> None:
 ERR_CANNOT_CONVERT_TO_ARROW_DOUBLE = 4037
 ERR_INVALID_INTEGER = 4038
 ERR_CANNOT_CONVERT_TO_ARROW_FLOAT = 4039
+ERR_ARROW_FIXED_SIZE_BINARY_VIOLATED = 4040
 
 # error numbers that result in InternalError
 ERR_MESSAGE_TYPE_UNKNOWN = 5000
@@ -575,6 +576,10 @@ def _raise_not_supported(feature: str) -> None:
         "Apache Arrow C Data structure overflow detected. A larger structure "
         "is needed."
     ),
+    ERR_ARROW_FIXED_SIZE_BINARY_VIOLATED: (
+        "value of length {actual_len} does not match the Apache Arrow fixed "
+        "size binary length of {fixed_size_len}"
+    ),
     ERR_ARROW_SPARSE_VECTOR_NOT_ALLOWED: (
         "Apache Arrow format does not support sparse vectors with flexible "
         "dimensions"

src/oracledb/impl/arrow/array.pyx

@@ -108,6 +108,11 @@ cdef class ArrowArrayImpl:
         cdef ArrowBufferView data
         data.data.data = ptr
         data.size_bytes = num_bytes
+        if self.schema_impl.fixed_size > 0 \
+                and num_bytes != self.schema_impl.fixed_size:
+            errors._raise_err(errors.ERR_ARROW_FIXED_SIZE_BINARY_VIOLATED,
+                              actual_len=num_bytes,
+                              fixed_size_len=self.schema_impl.fixed_size)
         _check_nanoarrow(ArrowArrayAppendBytes(self.arrow_array, data))
 
     cdef int append_decimal(self, void* ptr, int64_t num_bytes) except -1:

tests/test_9300_dataframe_requested_schema.py

@@ -745,3 +745,13 @@ def test_9326(dtype, value, conn, test_env):
         conn.fetch_df_all(
             "select :1 from dual", [value], requested_schema=requested_schema
         )
+
+
+@pytest.mark.parametrize("value", [b"Too short", b"Much too long"])
+def test_9327(value, conn, test_env):
+    "9327 - fetch_df_all() with fixed width binary violations"
+    requested_schema = pyarrow.schema([("VALUE", pyarrow.binary(length=10))])
+    with test_env.assert_raises_full_code("DPY-4040"):
+        conn.fetch_df_all(
+            "select :1 from dual", [value], requested_schema=requested_schema
+        )

tests/test_9400_dataframe_requested_schema_async.py

@@ -768,3 +768,13 @@ async def test_9426(dtype, value, async_conn, test_env):
         await async_conn.fetch_df_all(
             "select :1 from dual", [value], requested_schema=requested_schema
         )
+
+
+@pytest.mark.parametrize("value", [b"Too short", b"Much too long"])
+async def test_9427(value, async_conn, test_env):
+    "9427 - fetch_df_all() with fixed width binary violations"
+    requested_schema = pyarrow.schema([("VALUE", pyarrow.binary(length=10))])
+    with test_env.assert_raises_full_code("DPY-4040"):
+        await async_conn.fetch_df_all(
+            "select :1 from dual", [value], requested_schema=requested_schema
+        )