feat: implement basic parquet writer and add roundtrip tests

.gitignore

@@ -23,3 +23,6 @@ cmake-build-release/
 
 # intellij files
 .idea
+
+# vscode files
+.vscode

src/iceberg/CMakeLists.txt

@@ -115,7 +115,8 @@ if(ICEBERG_BUILD_BUNDLE)
       parquet/parquet_data_util.cc
       parquet/parquet_reader.cc
       parquet/parquet_register.cc
-      parquet/parquet_schema_util.cc)
+      parquet/parquet_schema_util.cc
+      parquet/parquet_writer.cc)
 
   # Libraries to link with exported libiceberg_bundle.{so,a}.
   set(ICEBERG_BUNDLE_STATIC_BUILD_INTERFACE_LIBS)

src/iceberg/arrow/arrow_error_transform_internal.h

@@ -30,6 +30,8 @@ inline ErrorKind ToErrorKind(const ::arrow::Status& status) {
   switch (status.code()) {
     case ::arrow::StatusCode::IOError:
       return ErrorKind::kIOError;
+    case ::arrow::StatusCode::NotImplemented:
+      return ErrorKind::kNotImplemented;
     default:
       return ErrorKind::kUnknownError;
   }

src/iceberg/constants.h

@@ -0,0 +1,28 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+#pragma once
+
+#include <string_view>
+
+namespace iceberg {
+
+constexpr std::string_view kParquetFieldIdKey = "PARQUET:field_id";
+
+}  // namespace iceberg

src/iceberg/parquet/parquet_reader.cc

@@ -125,9 +125,9 @@ class ParquetReader::Impl {
     arrow_reader_properties.set_arrow_extensions_enabled(true);
 
     // Open the Parquet file reader
-    ICEBERG_ASSIGN_OR_RAISE(auto input_stream, OpenInputStream(options));
+    ICEBERG_ASSIGN_OR_RAISE(input_stream_, OpenInputStream(options));
     auto file_reader =
-        ::parquet::ParquetFileReader::Open(std::move(input_stream), reader_properties);
+        ::parquet::ParquetFileReader::Open(input_stream_, reader_properties);
     ICEBERG_ARROW_RETURN_NOT_OK(::parquet::arrow::FileReader::Make(
         pool_, std::move(file_reader), arrow_reader_properties, &reader_));
 
@@ -169,6 +169,7 @@ class ParquetReader::Impl {
     }
 
     reader_.reset();
+    ICEBERG_ARROW_RETURN_NOT_OK(input_stream_->Close());
     return {};
   }
 
@@ -236,6 +237,8 @@ class ParquetReader::Impl {
   std::shared_ptr<::iceberg::Schema> read_schema_;
   // The projection result to apply to the read schema.
   SchemaProjection projection_;
+  // The input stream to read Parquet file.
+  std::shared_ptr<::arrow::io::RandomAccessFile> input_stream_;
   // Parquet file reader to create RecordBatchReader.
   std::unique_ptr<::parquet::arrow::FileReader> reader_;
   // The context to keep track of the reading progress.

src/iceberg/parquet/parquet_register.cc

@@ -21,8 +21,6 @@
 
 namespace iceberg::parquet {
 
-void RegisterWriter() {}
-
 void RegisterAll() {
   RegisterReader();
   RegisterWriter();

src/iceberg/parquet/parquet_schema_util.cc

@@ -26,6 +26,7 @@
 #include <parquet/arrow/schema.h>
 #include <parquet/schema.h>
 
+#include "iceberg/constants.h"
 #include "iceberg/metadata_columns.h"
 #include "iceberg/parquet/parquet_schema_util_internal.h"
 #include "iceberg/result.h"
@@ -38,8 +39,6 @@ namespace iceberg::parquet {
 
 namespace {
 
-constexpr std::string_view kParquetFieldIdKey = "PARQUET:field_id";
-
 std::optional<int32_t> FieldIdFromMetadata(
     const std::shared_ptr<const ::arrow::KeyValueMetadata>& metadata) {
   if (!metadata) {

src/iceberg/parquet/parquet_writer.cc

@@ -0,0 +1,166 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+#include "iceberg/parquet/parquet_writer.h"
+
+#include <memory>
+
+#include <arrow/c/bridge.h>
+#include <arrow/record_batch.h>
+#include <arrow/util/key_value_metadata.h>
+#include <parquet/arrow/schema.h>
+#include <parquet/arrow/writer.h>
+#include <parquet/file_writer.h>
+#include <parquet/properties.h>
+
+#include "iceberg/arrow/arrow_error_transform_internal.h"
+#include "iceberg/arrow/arrow_fs_file_io_internal.h"
+#include "iceberg/schema_internal.h"
+#include "iceberg/util/checked_cast.h"
+#include "iceberg/util/macros.h"
+
+namespace iceberg::parquet {
+
+namespace {
+
+Result<std::shared_ptr<::arrow::io::OutputStream>> OpenOutputStream(
+    const WriterOptions& options) {
+  auto io = internal::checked_pointer_cast<arrow::ArrowFileSystemFileIO>(options.io);
+  ICEBERG_ARROW_ASSIGN_OR_RETURN(auto output, io->fs()->OpenOutputStream(options.path));
+  return output;
+}
+
+}  // namespace
+
+class ParquetWriter::Impl {
+ public:
+  Status Open(const WriterOptions& options) {
+    auto writer_properties =
+        ::parquet::WriterProperties::Builder().memory_pool(pool_)->build();
+    auto arrow_writer_properties = ::parquet::default_arrow_writer_properties();
+
+    ArrowSchema c_schema;
+    ICEBERG_RETURN_UNEXPECTED(ToArrowSchema(*options.schema, &c_schema));
+    ICEBERG_ARROW_ASSIGN_OR_RETURN(arrow_schema_, ::arrow::ImportSchema(&c_schema));
+
+    std::shared_ptr<::parquet::SchemaDescriptor> schema_descriptor;
+    ICEBERG_ARROW_RETURN_NOT_OK(
+        ::parquet::arrow::ToParquetSchema(arrow_schema_.get(), *writer_properties,
+                                          *arrow_writer_properties, &schema_descriptor));
+    auto schema_node = std::static_pointer_cast<::parquet::schema::GroupNode>(
+        schema_descriptor->schema_root());
+
+    ICEBERG_ASSIGN_OR_RAISE(output_stream_, OpenOutputStream(options));
+    auto file_writer = ::parquet::ParquetFileWriter::Open(output_stream_, schema_node,
+                                                          writer_properties);
+    ICEBERG_ARROW_RETURN_NOT_OK(::parquet::arrow::FileWriter::Make(
+        pool_, std::move(file_writer), arrow_schema_, arrow_writer_properties, &writer_));
+
+    return {};
+  }
+
+  Status Write(ArrowArray array) {
+    ICEBERG_ARROW_ASSIGN_OR_RETURN(auto batch,
+                                   ::arrow::ImportRecordBatch(&array, arrow_schema_));
+
+    ICEBERG_ARROW_RETURN_NOT_OK(writer_->WriteRecordBatch(*batch));
+
+    return {};
+  }
+
+  // Close the writer and release resources
+  Status Close() {
+    if (writer_ == nullptr) {
+      return {};  // Already closed
+    }
+
+    ICEBERG_ARROW_RETURN_NOT_OK(writer_->Close());
+    auto& metadata = writer_->metadata();
+    split_offsets_.reserve(metadata->num_row_groups());
+    for (int i = 0; i < metadata->num_row_groups(); ++i) {
+      split_offsets_.push_back(metadata->RowGroup(i)->file_offset());
+    }
+    writer_.reset();
+
+    ICEBERG_ARROW_ASSIGN_OR_RETURN(total_bytes_, output_stream_->Tell());
+    ICEBERG_ARROW_RETURN_NOT_OK(output_stream_->Close());
+    return {};
+  }
+
+  bool Closed() const { return writer_ == nullptr; }
+
+  int64_t length() const { return total_bytes_; }
+
+  std::vector<int64_t> split_offsets() const { return split_offsets_; }
+
+ private:
+  // TODO(gangwu): make memory pool configurable
+  ::arrow::MemoryPool* pool_ = ::arrow::default_memory_pool();
+  // Schema to write from the Parquet file.
+  std::shared_ptr<::arrow::Schema> arrow_schema_;
+  // The output stream to write Parquet file.
+  std::shared_ptr<::arrow::io::OutputStream> output_stream_;
+  // Parquet file writer to write ArrowArray.
+  std::unique_ptr<::parquet::arrow::FileWriter> writer_;
+  // Total length of the written Parquet file.
+  int64_t total_bytes_;
+  // Row group start offsets in the Parquet file.
+  std::vector<int64_t> split_offsets_;
+};
+
+ParquetWriter::~ParquetWriter() = default;
+
+Status ParquetWriter::Open(const WriterOptions& options) {
+  impl_ = std::make_unique<Impl>();
+  return impl_->Open(options);
+}
+
+Status ParquetWriter::Write(ArrowArray array) { return impl_->Write(array); }
+
+Status ParquetWriter::Close() { return impl_->Close(); }
+
+std::optional<Metrics> ParquetWriter::metrics() {
+  if (!impl_->Closed()) {
+    return std::nullopt;
+  }
+  return {};
+}
+
+std::optional<int64_t> ParquetWriter::length() {
+  if (!impl_->Closed()) {
+    return std::nullopt;
+  }
+  return impl_->length();
+}
+
+std::vector<int64_t> ParquetWriter::split_offsets() {
+  if (!impl_->Closed()) {
+    return {};
+  }
+  return impl_->split_offsets();
+}
+
+void RegisterWriter() {
+  static WriterFactoryRegistry parquet_writer_register(
+      FileFormatType::kParquet, []() -> Result<std::unique_ptr<Writer>> {
+        return std::make_unique<ParquetWriter>();
+      });
+}
+
+}  // namespace iceberg::parquet

src/iceberg/parquet/parquet_writer.h

@@ -0,0 +1,51 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+#pragma once
+
+#include "iceberg/file_writer.h"
+#include "iceberg/iceberg_bundle_export.h"
+
+namespace iceberg::parquet {
+
+/// \brief A writer that writes ArrowArray to Parquet files.
+class ICEBERG_BUNDLE_EXPORT ParquetWriter : public Writer {
+ public:
+  ParquetWriter() = default;
+
+  ~ParquetWriter() override;
+
+  Status Open(const WriterOptions& options) final;
+
+  Status Close() final;
+
+  Status Write(ArrowArray array) final;
+
+  std::optional<Metrics> metrics() final;
+
+  std::optional<int64_t> length() final;
+
+  std::vector<int64_t> split_offsets() final;
+
+ private:
+  class Impl;
+  std::unique_ptr<Impl> impl_;
+};
+
+}  // namespace iceberg::parquet

src/iceberg/schema_internal.cc

@@ -24,6 +24,7 @@
 #include <optional>
 #include <string>
 
+#include "iceberg/constants.h"
 #include "iceberg/schema.h"
 #include "iceberg/type.h"
 #include "iceberg/util/macros.h"
@@ -45,7 +46,7 @@ ArrowErrorCode ToArrowSchema(const Type& type, bool optional, std::string_view n
   NANOARROW_RETURN_NOT_OK(ArrowMetadataBuilderInit(&metadata_buffer, nullptr));
   if (field_id.has_value()) {
     NANOARROW_RETURN_NOT_OK(ArrowMetadataBuilderAppend(
-        &metadata_buffer, ArrowCharView(std::string(kFieldIdKey).c_str()),
+        &metadata_buffer, ArrowCharView(std::string(kParquetFieldIdKey).c_str()),
         ArrowCharView(std::to_string(field_id.value()).c_str())));
   }
 
@@ -185,8 +186,8 @@ int32_t GetFieldId(const ArrowSchema& schema) {
     return kUnknownFieldId;
   }
 
-  ArrowStringView field_id_key{.data = kFieldIdKey.data(),
-                               .size_bytes = kFieldIdKey.size()};
+  ArrowStringView field_id_key{.data = kParquetFieldIdKey.data(),
+                               .size_bytes = kParquetFieldIdKey.size()};
   ArrowStringView field_id_value;
   if (ArrowMetadataGetValue(schema.metadata, field_id_key, &field_id_value) !=
       NANOARROW_OK) {

src/iceberg/schema_internal.h

@@ -30,11 +30,6 @@
 
 namespace iceberg {
 
-// Apache Arrow C++ uses "PARQUET:field_id" to store field IDs for Parquet.
-// Here we follow a similar convention for Iceberg but we might also add
-// "PARQUET:field_id" in the future once we implement a Parquet writer.
-constexpr std::string_view kFieldIdKey = "ICEBERG:field_id";
-
 /// \brief Convert an Iceberg schema to an Arrow schema.
 ///
 /// \param[in] schema The Iceberg schema to convert.

src/iceberg/type.cc

@@ -319,4 +319,16 @@ std::shared_ptr<FixedType> fixed(int32_t length) {
   return std::make_shared<FixedType>(length);
 }
 
+std::shared_ptr<MapType> map(SchemaField key, SchemaField value) {
+  return std::make_shared<MapType>(key, value);
+}
+
+std::shared_ptr<ListType> list(SchemaField element) {
+  return std::make_shared<ListType>(std::move(element));
+}
+
+std::shared_ptr<StructType> struct_(std::vector<SchemaField> fields) {
+  return std::make_shared<StructType>(std::move(fields));
+}
+
 }  // namespace iceberg