fix metadata size

Author: Alex-PLACET

include/sparrow_ipc/stream_file_serializer.hpp

@@ -202,7 +202,7 @@ namespace sparrow_ipc
                 // Serialize and get block info
                 const auto info = serialize_record_batch(rb, m_stream, m_compression);
 
-                m_record_batch_blocks.push_back({offset, info.metadata_length, info.body_length});
+                m_record_batch_blocks.emplace_back(offset, info.metadata_length, info.body_length);
             }
         }
 

src/serialize.cpp

@@ -1,19 +1,22 @@
+#include "sparrow_ipc/serialize.hpp"
+
+#include <cstdint>
 #include <optional>
 
-#include "sparrow_ipc/serialize.hpp"
 #include "sparrow_ipc/flatbuffer_utils.hpp"
 #include "sparrow_ipc/utils.hpp"
 
 namespace sparrow_ipc
 {
-    void common_serialize(
-        const flatbuffers::FlatBufferBuilder& builder,
-        any_output_stream& stream
-    )
+    void common_serialize(const flatbuffers::FlatBufferBuilder& builder, any_output_stream& stream)
     {
         stream.write(continuation);
         const flatbuffers::uoffset_t size = builder.GetSize();
-        const std::span<const uint8_t> size_span(reinterpret_cast<const uint8_t*>(&size), sizeof(uint32_t));
+        const int32_t size_with_padding = utils::align_to_8(static_cast<int32_t>(size));
+        const std::span<const uint8_t> size_span(
+            reinterpret_cast<const uint8_t*>(&size_with_padding),
+            sizeof(int32_t)
+        );
         stream.write(size_span);
         stream.write(std::span(builder.GetBufferPointer(), size));
         stream.add_padding();
@@ -24,39 +27,27 @@ namespace sparrow_ipc
         common_serialize(get_schema_message_builder(record_batch), stream);
     }
 
-    serialized_record_batch_info serialize_record_batch(const sparrow::record_batch& record_batch, any_output_stream& stream, std::optional<CompressionType> compression)
+    serialized_record_batch_info serialize_record_batch(
+        const sparrow::record_batch& record_batch,
+        any_output_stream& stream,
+        std::optional<CompressionType> compression
+    )
     {
         // Build and serialize metadata
         flatbuffers::FlatBufferBuilder builder = get_record_batch_message_builder(record_batch, compression);
-        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);
-        
+
         // Track position before body to calculate body length
         const size_t body_start = stream.size();
-        
+
         // Write body
         generate_body(record_batch, stream, compression);
-        
-        // Calculate body length (should already be 8-aligned since generate_body pads each buffer)
-        const int64_t body_length = static_cast<int64_t>(stream.size() - body_start);
-        
-        return {metadata_length, body_length};
+
+        const auto body_length = static_cast<int64_t>(stream.size() - body_start);
+        const flatbuffers::uoffset_t flatbuffer_size = builder.GetSize();
+        const auto metadata_length = static_cast<int32_t>(utils::align_to_8(flatbuffer_size));
+        return {.metadata_length = metadata_length, .body_length = body_length};
     }
 }