wip

Author: Alex-PLACET

include/sparrow_ipc/serialize.hpp

@@ -89,7 +89,7 @@ namespace sparrow_ipc
      *       includes both metadata and data portions of the record batch
      */
 
-    SPARROW_IPC_API void
+    SPARROW_IPC_API serialized_record_batch_info
     serialize_record_batch(const sparrow::record_batch& record_batch,
                            any_output_stream& stream,
                            std::optional<CompressionType> compression,

include/sparrow_ipc/serialize_utils.hpp

@@ -24,31 +24,7 @@ namespace sparrow_ipc
      */
     SPARROW_IPC_API void
     serialize_schema_message(const sparrow::record_batch& record_batch, any_output_stream& stream);
-
-    /**
-     * @brief Serializes a record batch into a binary format following the Arrow IPC specification.
-     *
-     * This function converts a sparrow record batch into a serialized byte vector that includes:
-     * - A continuation marker at the beginning
-     * - The record batch metadata length (4 bytes)
-     * - The FlatBuffer-encoded record batch metadata containing field nodes and buffer information
-     * - Padding to ensure 8-byte alignment
-     * - The actual data body containing the record batch buffers
-     *
-     * The serialization follows the Arrow IPC stream format where each record batch message
-     * consists of a metadata section followed by a body section containing the actual data.
-     *
-     * @param record_batch The sparrow record batch to be serialized.
-     * @param stream The output stream where the serialized record batch will be written.
-     * @param compression Optional: The compression type to use when serializing.
-     * @param cache Optional: A cache for compressed buffers to avoid recompression if compression is enabled.
-     * If compression is given, cache should be set as well.
-     */
-    SPARROW_IPC_API void
-    serialize_record_batch(const sparrow::record_batch& record_batch, any_output_stream& stream,
-                           std::optional<CompressionType> compression,
-                           std::optional<std::reference_wrapper<CompressionCache>> cache);
-
+    
     /**
      * @brief Calculates the total serialized size of a schema message.
      *

src/serialize.cpp

@@ -30,29 +30,23 @@ namespace sparrow_ipc
     serialized_record_batch_info serialize_record_batch(
         const sparrow::record_batch& record_batch,
         any_output_stream& stream,
-       
-                                std::optional<CompressionType> compression
-    ,
-                                std::optional<std::reference_wrapper<CompressionCache>> cache)
+        std::optional<CompressionType> compression,
+        std::optional<std::reference_wrapper<CompressionCache>> cache
+    )
     {
         // Build and serialize metadata
         flatbuffers::FlatBufferBuilder builder = get_record_batch_message_builder(record_batch, compression, cache);
-        const flatbuffers::uoffset_t flatbuffer_size = builder.GetSize();
-        
+
         // Calculate metadata length for the Block in the footer
         // According to Arrow spec, metadata_length must be a multiple of 8.
         // The encapsulated message format is:
         //   - continuation (4 bytes)
         //   - size prefix (4 bytes)
         //   - flatbuffer metadata (flatbuffer_size bytes)
         //   - padding to 8-byte boundary
-        // 
+        //
         // Arrow's WriteMessage returns metadata_length = align_to_8(8 + flatbuffer_size)
         // which INCLUDES the continuation bytes.
-        const size_t prefix_size = continuation.size() + sizeof(uint32_t);  // 8 bytes
-        const int32_t metadata_length = static_cast<int32_t>(
-            utils::align_to_8(prefix_size + flatbuffer_size)
-        );
 
         // Write metadata
         common_serialize(builder, stream);