GH-15256: [C++][Dataset] Add support for writing with Partitioning::Default()

c_glib/arrow-dataset-glib/partitioning.cpp

@@ -38,9 +38,6 @@ G_BEGIN_DECLS
  * #GADatasetPartitioning is a base class for partitioning classes
  * such as #GADatasetDirectoryPartitioning.
  *
- * #GADatasetDefaultPartitioning is a class for partitioning that
- * doesn't partition.
- *
  * #GADatasetKeyValuePartitioningOptions is a class for key-value
  * partitioning options.
  *
@@ -345,35 +342,19 @@ gadataset_partitioning_get_type_name(GADatasetPartitioning *partitioning)
 }
 
 
-G_DEFINE_TYPE(GADatasetDefaultPartitioning,
- gadataset_default_partitioning,
- GADATASET_TYPE_PARTITIONING)
-
-static void
-gadataset_default_partitioning_init(GADatasetDefaultPartitioning *object)
-{
-}
-
-static void
-gadataset_default_partitioning_class_init(
- GADatasetDefaultPartitioningClass *klass)
-{
-}
-
 /**
- * gadataset_default_partitioning_new:
+ * gadataset_partitioning_create_default:
  *
- * Returns: The newly created #GADatasetDefaultPartitioning that
- * doesn't partition.
+ * Returns: (transfer full): The newly created #GADatasetPartitioning
+ * that doesn't partition.
  *
- * Since: 11.0.0
+ * Since: 12.0.0
  */
-GADatasetDefaultPartitioning *
-gadataset_default_partitioning_new(void)
+GADatasetPartitioning *
+gadataset_partitioning_create_default(void)
 {
  auto arrow_partitioning = arrow::dataset::Partitioning::Default();
- return GADATASET_DEFAULT_PARTITIONING(
- gadataset_partitioning_new_raw(&arrow_partitioning));
+ return gadataset_partitioning_new_raw(&arrow_partitioning);
 }
 
 
@@ -813,9 +794,7 @@ gadataset_partitioning_new_raw(
 {
  GType type = GADATASET_TYPE_PARTITIONING;
  const auto arrow_type_name = (*arrow_partitioning)->type_name();
- if (arrow_type_name == "default") {
- type = GADATASET_TYPE_DEFAULT_PARTITIONING;
- } else if (arrow_type_name == "directory") {
+ if (arrow_type_name == "directory") {
  type = GADATASET_TYPE_DIRECTORY_PARTITIONING;
  } else if (arrow_type_name == "hive") {
  type = GADATASET_TYPE_HIVE_PARTITIONING;

c_glib/arrow-dataset-glib/partitioning.h

@@ -71,21 +71,9 @@ gchar *
 gadataset_partitioning_get_type_name(GADatasetPartitioning *partitioning);
 
 
-#define GADATASET_TYPE_DEFAULT_PARTITIONING \
- (gadataset_default_partitioning_get_type())
-G_DECLARE_DERIVABLE_TYPE(GADatasetDefaultPartitioning,
- gadataset_default_partitioning,
- GADATASET,
- DEFAULT_PARTITIONING,
- GADatasetPartitioning)
-struct _GADatasetDefaultPartitioningClass
-{
- GADatasetPartitioningClass parent_class;
-};
-
-GARROW_AVAILABLE_IN_11_0
-GADatasetDefaultPartitioning *
-gadataset_default_partitioning_new(void);
+GARROW_AVAILABLE_IN_12_0
+GADatasetPartitioning *
+gadataset_partitioning_create_default(void);
 
 
 #define GADATASET_TYPE_KEY_VALUE_PARTITIONING_OPTIONS \

c_glib/test/dataset/test-partitioning.rb

@@ -23,7 +23,8 @@ def setup
  end
 
  def test_default
- assert_equal("default", ArrowDataset::DefaultPartitioning.new.type_name)
+ assert_equal("directory",
+ ArrowDataset::Partitioning.create_default.type_name)
  end
 
  def test_directory

cpp/src/arrow/dataset/partition.cc

@@ -75,32 +75,7 @@ std::string StripNonPrefix(const std::string& path) {
 } // namespace
 
 std::shared_ptr<Partitioning> Partitioning::Default() {
- class DefaultPartitioning : public Partitioning {
- public:
- DefaultPartitioning() : Partitioning(::arrow::schema({})) {}
-
- std::string type_name() const override { return "default"; }
-
- bool Equals(const Partitioning& other) const override {
- return type_name() == other.type_name();
- }
-
- Result<compute::Expression> Parse(const std::string& path) const override {
- return compute::literal(true);
- }
-
- Result<PartitionPathFormat> Format(const compute::Expression& expr) const override {
- return Status::NotImplemented("formatting paths from ", type_name(),
- " Partitioning");
- }
-
- Result<PartitionedBatches> Partition(
- const std::shared_ptr<RecordBatch>& batch) const override {
- return PartitionedBatches{{batch}, {compute::literal(true)}};
- }
- };
-
- return std::make_shared<DefaultPartitioning>();
+ return std::make_shared<DirectoryPartitioning>(arrow::schema({}));
 }
 
 static Result<RecordBatchVector> ApplyGroupings(

cpp/src/arrow/dataset/partition.h

@@ -90,7 +90,8 @@ class ARROW_DS_EXPORT Partitioning : public util::EqualityComparable<Partitionin
 
  virtual Result<PartitionPathFormat> Format(const compute::Expression& expr) const = 0;
 
- /// \brief A default Partitioning which always yields scalar(true)
+ /// \brief A default Partitioning which is a DirectoryPartitioning
+ /// with an empty schema.
  static std::shared_ptr<Partitioning> Default();
 
  /// \brief The partition schema.

cpp/src/arrow/dataset/partition_test.cc

@@ -189,6 +189,12 @@ TEST_F(TestPartitioning, Partition) {
  expected_expressions);
 }
 
+TEST_F(TestPartitioning, DefaultPartitioningIsDirectoryPartitioning) {
+ auto partitioning = Partitioning::Default();
+ ASSERT_EQ(partitioning->type_name(), "directory");
+ AssertSchemaEqual(partitioning->schema(), schema({}));
+}
+
 TEST_F(TestPartitioning, DirectoryPartitioning) {
  partitioning_ = std::make_shared<DirectoryPartitioning>(
  schema({field("alpha", int32()), field("beta", utf8())}));
@@ -209,6 +215,18 @@ TEST_F(TestPartitioning, DirectoryPartitioning) {
  equal(field_ref("beta"), literal("foo"))));
 }
 
+TEST_F(TestPartitioning, DirectoryPartitioningEmpty) {
+ partitioning_ = std::make_shared<DirectoryPartitioning>(schema({}));
+ written_schema_ = partitioning_->schema();
+
+ // No partitioning info
+ AssertParse("", literal(true));
+ // Files can be in subdirectories
+ AssertParse("/foo/", literal(true));
+ // Partitioning info is discarded on write
+ AssertFormat(equal(field_ref("alpha"), literal(7)), "");
+}
+
 TEST_F(TestPartitioning, DirectoryPartitioningEquals) {
  auto part = std::make_shared<DirectoryPartitioning>(
  schema({field("alpha", int32()), field("beta", utf8())}));

python/pyarrow/tests/test_dataset.py

@@ -3640,12 +3640,12 @@ def test_dataset_preserved_partitioning(tempdir):
  # through discovery, but without partitioning
  _, path = _create_single_file(tempdir)
  dataset = ds.dataset(path)
- assert dataset.partitioning is None
+ assert isinstance(dataset.partitioning, ds.DirectoryPartitioning)
 
  # through discovery, with hive partitioning but not specified
  full_table, path = _create_partitioned_dataset(tempdir)
  dataset = ds.dataset(path)
- assert dataset.partitioning is None
+ assert isinstance(dataset.partitioning, ds.DirectoryPartitioning)
 
  # through discovery, with hive partitioning (from a partitioning factory)
  dataset = ds.dataset(path, partitioning="hive")