I have a List<object>
with different types of objects in it like integers, strings, and custom types. All custom types are protobuf-adjusted.
What I wanna do now is to serialize / deserialize this list with protobuf.net. Up until now I suspect that I have to declare each and every type explicitly, which is unfortunately not possible with these mixed-list constructs. Because the binary formater has no problems to do these things I hope that I missed something and that you can help me out.
So my question is how to deal with objects in protobuf.net.

- 1,026,079
- 266
- 2,566
- 2,900
-
please also tell me that how much data consumption will be there to serialize a list having more than 10,000 element as an example? – LoveToCode Aug 23 '16 at 11:46
3 Answers
(disclosure: I'm the author of protobuf-net)
BinaryFormatter
is a metadata-based serializer; i.e. it sends .NET type information about every object serialized. protobuf-net is a contract-based serializer (the binary equivalent of XmlSerializer
/ DataContractSerializer
, which will also reject this).
There is no current mechanism for transporting arbitrary objects, since the other end will have no way of knowing what you are sending; however, if you have a known set of different object types you want to send, there may be options. There is also work in the pipeline to allow runtime-extensible schemas (rather than just attributes, which are fixed at build) - but this is far from complete.
This isn't ideal, but it works... it should be easier when I've completed the work to support runtime schemas:
using System;
using System.Collections.Generic;
using ProtoBuf;
[ProtoContract]
[ProtoInclude(10, typeof(DataItem<int>))]
[ProtoInclude(11, typeof(DataItem<string>))]
[ProtoInclude(12, typeof(DataItem<DateTime>))]
[ProtoInclude(13, typeof(DataItem<Foo>))]
abstract class DataItem {
public static DataItem<T> Create<T>(T value) {
return new DataItem<T>(value);
}
public object Value {
get { return ValueImpl; }
set { ValueImpl = value; }
}
protected abstract object ValueImpl {get;set;}
protected DataItem() { }
}
[ProtoContract]
sealed class DataItem<T> : DataItem {
public DataItem() { }
public DataItem(T value) { Value = value; }
[ProtoMember(1)]
public new T Value { get; set; }
protected override object ValueImpl {
get { return Value; }
set { Value = (T)value; }
}
}
[ProtoContract]
public class Foo {
[ProtoMember(1)]
public string Bar { get; set; }
public override string ToString() {
return "Foo with Bar=" + Bar;
}
}
static class Program {
static void Main() {
var items = new List<DataItem>();
items.Add(DataItem.Create(12345));
items.Add(DataItem.Create(DateTime.Today));
items.Add(DataItem.Create("abcde"));
items.Add(DataItem.Create(new Foo { Bar = "Marc" }));
items.Add(DataItem.Create(67890));
// serialize and deserialize
var clone = Serializer.DeepClone(items);
foreach (DataItem item in clone) {
Console.WriteLine(item.Value);
}
}
}

- 1,026,079
- 266
- 2,566
- 2,900
-
Thanks Gravell, actualy I realy like your work and hoped for an answer directly from you. I know which datatypes are in the list so I'm really curious about your solution. – May 29 '09 at 07:38
-
Thanks Gravell, can you give an estimate when you'll have finished the support of runtime schemas? – May 29 '09 at 12:17
-
I'd rather not; I gave an estimate before, and it didn't fit... there are a lot of edge-cases to fit, even *before* I optimise it. If I had to guess, I reckon that unless I get a good block of free time (which I don't expect) it'll take another couple of months. I'm also trying to keep the "trunk" up to date with the latest wire-format changes, which makes things even more fun. – Marc Gravell May 29 '09 at 12:59
List<YourClass> list;
ProtoBuf.Serializer.Deserialize<List<YourClass>>(filestream);
There is a way of doing this, albeit not a very clean way, by using a wrapper object that utilises another serialisation mechanism that supports arbitrary objects. I am presenting an example below using JSON but, like I said, resorting to a different serialisation tool seems like defeating the purpose of using protobuf:
[DataContract]
public class ObjectWrapper
{
[DataMember(Order = 1)]
private readonly string _serialisedContent;
[DataMember(Order = 2)]
private readonly string _serialisedType;
public object Content { get; private set; }
[UsedImplicitly]
private ObjectWrapper() { }
public ObjectWrapper(object content)
{
_serialisedContent = JsonConvert.SerializeObject(content);
_serialisedType = content.GetType().FullName;
Content = content;
}
[ProtoAfterDeserialization]
private void Initialise()
{
var type = Type.GetType(_serialisedType);
Content = type != null
? JsonConvert.DeserializeObject(_serialisedContent, type)
: JsonConvert.DeserializeObject(_serialisedContent);
}
}
EDIT: This can also be done using C#'s built-in binary serialisation
[DataContract]
public class ObjectWrapper
{
[DataMember(Order = 1)]
private readonly string _serialisedContent;
public object Content { get; private set; }
[UsedImplicitly]
private ObjectWrapper() { }
public ObjectWrapper(object content)
{
using (var stream = new MemoryStream())
{
var formatter = new BinaryFormatter();
formatter.Serialize(stream, content);
stream.Flush();
stream.Position = 0;
_serialisedContent = Convert.ToBase64String(stream.ToArray());
}
}
[ProtoAfterDeserialization]
private void Initialise()
{
var data = Convert.FromBase64String(source);
using (var stream = new MemoryStream(data))
{
var formatter = new BinaryFormatter();
stream.Seek(0, SeekOrigin.Begin);
return formatter.Deserialize(stream);
}
}
}

- 886
- 1
- 10
- 26