ARROW-8070: [C++] Cast segfaults on unsupported cast from list<binary…

…> to utf8 Return with ArrowInvalid for `list<...> => not list` casts. The `list(binary()) => utf8()` could make sense, but as a `concat` operation. Closes #6738 from kszucs/ARROW-8070 Authored-by: Krisztián Szűcs <szucs.krisztian@gmail.com> Signed-off-by: Wes McKinney <wesm+git@apache.org>
apache · Mar 27, 2020 · 8ee4be7 · 8ee4be7
1 parent 26bd82f
commit 8ee4be7
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 9 deletions.
diff --git a/cpp/src/arrow/compute/kernels/cast.cc b/cpp/src/arrow/compute/kernels/cast.cc
@@ -1429,8 +1429,8 @@ Status GetListCastFunc(const DataType& in_type, std::shared_ptr<DataType> out_ty
                        const CastOptions& options,
                        std::unique_ptr<CastKernelBase>* kernel) {
   if (out_type->id() != TypeClass::type_id) {
-    // Kernel will be null
-    return Status::OK();
+    return Status::Invalid("Cannot cast from ", in_type.ToString(), " to ",
+                           out_type->ToString());
   }
   const DataType& in_value_type = *checked_cast<const TypeClass&>(in_type).value_type();
   std::shared_ptr<DataType> out_value_type =
@@ -1503,19 +1503,18 @@ Status GetCastFunction(const DataType& in_type, std::shared_ptr<DataType> out_ty
     CAST_FUNCTION_CASE(LargeStringType);
     CAST_FUNCTION_CASE(DictionaryType);
     case Type::NA:
-      cast_kernel.reset(new FromNullCastKernel(std::move(out_type)));
+      cast_kernel.reset(new FromNullCastKernel(out_type));
       break;
     case Type::LIST:
-      RETURN_NOT_OK(
-          GetListCastFunc<ListType>(in_type, std::move(out_type), options, &cast_kernel));
+      RETURN_NOT_OK(GetListCastFunc<ListType>(in_type, out_type, options, &cast_kernel));
       break;
     case Type::LARGE_LIST:
-      RETURN_NOT_OK(GetListCastFunc<LargeListType>(in_type, std::move(out_type), options,
-                                                   &cast_kernel));
+      RETURN_NOT_OK(
+          GetListCastFunc<LargeListType>(in_type, out_type, options, &cast_kernel));
       break;
     case Type::EXTENSION:
-      RETURN_NOT_OK(ExtensionCastKernel::Make(std::move(in_type), std::move(out_type),
-                                              options, &cast_kernel));
+      RETURN_NOT_OK(
+          ExtensionCastKernel::Make(std::move(in_type), out_type, options, &cast_kernel));
       break;
     default:
       break;

diff --git a/cpp/src/arrow/compute/kernels/cast_test.cc b/cpp/src/arrow/compute/kernels/cast_test.cc
@@ -1425,6 +1425,17 @@ TEST_F(TestCast, BooleanToString) { TestCastBooleanToString<StringType>(); }
 
 TEST_F(TestCast, BooleanToLargeString) { TestCastBooleanToString<LargeStringType>(); }
 
+TEST_F(TestCast, ListToPrimitive) {
+  auto from_int = ArrayFromJSON(list(int8()), "[[1, 2], [3, 4]]");
+  auto from_binary = ArrayFromJSON(list(binary()), "[[\"1\", \"2\"], [\"3\", \"4\"]]");
+
+  CastOptions options;
+  std::shared_ptr<Array> result;
+
+  ASSERT_RAISES(Invalid, Cast(&ctx_, *from_int, uint8(), options, &result));
+  ASSERT_RAISES(Invalid, Cast(&ctx_, *from_binary, utf8(), options, &result));
+}
+
 TEST_F(TestCast, ListToList) {
   CastOptions options;
   std::shared_ptr<Array> offsets;

diff --git a/python/pyarrow/tests/test_array.py b/python/pyarrow/tests/test_array.py
@@ -897,6 +897,17 @@ def test_cast_none():
         arr.cast(None)
 
 
+def test_cast_list_to_primitive():
+    # ARROW-8070: cast segfaults on unsupported cast from list<binary> to utf8
+    arr = pa.array([[1, 2], [3, 4]])
+    with pytest.raises(pa.ArrowInvalid):
+        arr.cast(pa.int8())
+
+    arr = pa.array([[b"a", b"b"], [b"c"]], pa.list_(pa.binary()))
+    with pytest.raises(pa.ArrowInvalid):
+        arr.cast(pa.binary())
+
+
 def test_cast_chunked_array():
     arrays = [pa.array([1, 2, 3]), pa.array([4, 5, 6])]
     carr = pa.chunked_array(arrays)