Friday, February 25, 2011

Managing Decoupling Part 3 - C++ Duck Typing

Some systems need to manipulate objects whose exact nature are not known. For example, a particle system has to manipulate particles that sometimes have mass, sometimes a full 3D rotation, sometimes only 2D rotation, etc. (A good particle system anyway, a bad particle system could use the same struct for all particles in all effects. And the struct could have some fields called custom_1,custom_2 used for different purposes in different effects. And it would be both inefficient, inflexible and messy.)

Another example is a networking system tasked with synchronizing game objects between clients and servers. A very general such system might want to treat the objects as open JSON-like structs, with arbitrary fields and values:

    "score" : 100,
    "name": "Player 1"

We want to be able to handle such “general” or “open” objects in C++ in a nice way. Since we care about structure we don’t want the system to be strongly coupled to the layout of the objects it manages. And since we are performance junkies, we would like to do it in a way that doesn’t completely kill performance. I.e., we don’t want everything to inherit from a base class Object and define our JSON-like objects as:

typedef std::map OpenStruct;

Generally speaking, there are three possible levels of flexibility with which we can work with objects and types in a programming language:

1. Exact typing - Only ducks are ducks

We require the object to be of a specific type. This is the typing method used in C and for classes without inheritance in C++.

2. Interface typing - If it says it’s a duck

We require the object to inherit from and implement a specific interface type. This is the typing method used by default in Java and C# and in C++ when inheritance and virtual methods are used. It is more flexible that the exact approach, but still introduces a coupling, because it forces the objects we manage to inherit a type defined by us.

Side rant: My general opinion is that while inheriting interfaces (abstract classes) is a valid and useful design tool, inheriting implementations is usually little more than a glorified “hack”, a way of patching parent classes by inserting custom code here and there. You almost always get a cleaner design when you build your objects with composition instead of with implementation inheritance.

3. Duck typing - If it quacks like a duck

We don’t care about the type of the object at all, as long as it has the fields and methods that we need. An example:

      def integrate_position(o, dt):
          o.position = o.position + o.velocity * dt

This method integrates the position of the object o. It doesn’t care what the type of o is, as long as it has a “position” field and a “velocity” field.

Duck typing is the default in many “scripting” languages such as Ruby, Python, Lua and JavaScript. The reflection interface of Java and C# can also be used for duck typing, but unfortunately the code tends to become far less elegant than in the scripting languages:

      o.GetType().GetProperty(“Position”).SetValue(o, o.GetType().
         GetProperty(“Position”).GetValue(o, null) + o.GetType().
         GetProperty(“Velocity”).GetValue(o, null) * dt, null)

What we want is some way of doing “duck typing” in C++.

Let’s look at inheritance and virtual functions first, since that is the standard way of “generalizing” code in C++. It is true that you could do general objects using the inheritance mechanism. You would create a class structure looking something like:

class Object {...};
class Int : public Object {...};
class Float : public Object{...};

and then use dynamic_cast or perhaps your own hand-rolled RTTI system to determine an object’s class.
But there are a number of drawbacks with this approach. It is quite verbose. The virtual inheritance model requires objects to be treated as pointers so they (probably) have to be heap allocated. This makes it tricky to get a good memory layout. And that hurts performance. Also, they are not PODs so we will have to do extra work if we want to move them to a co-processor or save them to disk.

So I prefer something much simpler. A generic object is just a type enum followed by the data for the object:

To pass the object you just pass its pointer. To make a copy, you make a copy of the memory block. You can also write it straight to disk and read it back, send it over network or to an SPU for off-core processing.

To extract the data from the object you would do something like:

unsigned type = *(unsigned *)o;
if (type == FLOAT_TYPE)
    float f = *(float *)(o + 4);

You don’t really need that many different object types: boolintfloatvector3quaternionstring,array and dictionary is usually enough. You can build more complicated types as aggregates of those, just as you do in JSON.

For a dictionary object we just store the name/key and type of each object:

I tend to use a four byte value for the name/key and not care if it is an integer, float or a 32-bit string hash. As long as the data is queried with the same key that it was stored with, the right value will be returned. I only use this method for small structs, so the probability for a hash collision is close to zero and can be handled by “manual resolution”.

If we have many objects with the same “dictionary type” (i.e. the same set of fields, just different values) it makes sense to break out the definition of the type from the data itself to save space:

Here the offset field stores the offset of each field in the data block. Now we can efficiently store an array of such data objects with just one copy of the dictionary type information:

Note that the storage space (and thereby the cache and memory performance) is exactly the same as if we were using an array of regular C structs, even though we are using a completely open free form JSON-like struct. And extracting or changing data just requires a little pointer arithmetic and a cast.

This would be a good way of storing particles in a particle system. (Note: This is an array-of-structures approach, you can of course also use duck typing with a sturcture-of-arrays approach. I leave that as an exercise to the reader.)

If you are a graphics programmer all of this should look pretty familiar. The “dictionary type description” is very much like a “vertex data description” and the “dictionary data” is awfully similar to “vertex data”. This should come as no big surprise. Vertex data is generic flexible data that needs to be processed fast in parallel on in-order processing units. It is not strange that with the same design criterions we end up with a similar solution.

Morale and musings

It is OK to manipulate blocks of raw memory! Pointer arithmetic does not destroy your program! Type casts are not “dirty”! Let your freak flag fly!

Data-oriented-design and object-oriented design are not polar opposites. As this example shows a data-oriented design can in a sense be “more object-oriented” than a standard C++ virtual function design, i.e., more similar to how objects work in high level languages such as Ruby and Lua.

On the other hand, data-oriented-design and inheritance are enemies. Because designs based on base class pointers and virtual functions want objects to live individually allocated on the heap. Which means you cannot control the memory layout. Which is what DOD is all about. (Yes, you can probably do clever tricks with custom allocators and patching of vtables for moving or deserializing objects, but why bother, DOD is simpler.)

You could also store function pointers in these open structs. Then you would have something very similar to Ruby/Lua objects. This could probably be used for something great. This is left as an exercise to the reader.


  1. Niklas, looks like you missed the title.

    Anyways, great stuff. Really digging your DOD approach.

  2. Niklas, this is another thought-provoking post.

    One thing that I'm struggling with is to work out how your json-like generic compressed (or at least compacted) data structure related to duck-typing. How to you see it being used to implement integration_position?

    Also, wanted to point out that this kind of thing is called an algebraic data type in ML-like languages.

    I typed up your example in Standard ML. It doesn't look like your trick of "optimized dictionaries" can be done in SML without resorting to similar low-level techniques that you have used in C++. Neither does the example guarantee a compact representation in continuous memory.

    This is similar to a JSON implementation that I was playing with recently as I learned Haskell:

  3. The way I see it, the data structure can be used to represent a generic object (similar to o in the Python example). We can implement something like o.position by scanning the object for a field called "position" and returning its value.

    I've done a bit of ML, but it was a long time ago. I think one of the problems with many functional languages is that they don't give you bit-level control of memory layout. A big disadvantage nowadays when memory performance is often the bottleneck. It would be interesting with a functional language that still allowed C-like control of data layout. I'm sure it exists somewhere :)

  4. What is the advantage of your typed data compared to having just a plain struct serialized? I see that you can change the layout of the data in memory freely as you need to query first where the field "position" is stored relative to the entry, but does this buy you that much in practice? Especially as you need to know that it's called position anyway (and not pos, for instance.)

    I somehow fail to see what this buys you besides some very ugly code with lots of type-casting back and forth.

    I wrote a small prototype implementation of a system as you described here and I'm curious to see where its strengths are. So far, I'm using "normal serialization" (i.e. structs have a load/save method; loading the fields individually) or plain memdumps (BVH for instance is offset based, storing/loading it as a single memory chunk.)

    Oh, and how do you create that data? The static data is easy to use, but generating requires to serialize part-by-part; especially if you do the nesting thing, handling all the temporary buffers seems to be a bit messy. I.e. if I want to create a dictionary of dictionaries, I need either a memory-hungry run-time representation of all of this or I serialise the nested dictionary somewhere else and then copy it (cheaply) around.

  5. You can write tools/functions that deal with objects without knowing the exact data layout. I.e., the tool can deal with anything that has a "position" or "pos" field.

    You can add fields dynamically... if you have something representing enemy data and a certain enemy needs a special field "time_until_rage". Then you don't have to create a special struct for that enemy, you can just add that field to the struct.

    If you haven't spent some time in a language where duck typing is the norm, such as Ruby and Lua, you should try it. It is a different programming style with its own specific advantages.

  6. Hello Niklas. I really like your articles, they offer a lot of interesting ideas on how to better ones architecture. Not just for game development but for development in general.

    But I'm having some trouble understanding how you implemented Duck Typing. Is it some kind of an array you are using? I know that you can move with the pointer by simply adding position, but how are you creating the generic object?

  7. Hi, I know this is really old post, but pretty interesting one. Is it possible to fix missing images ? thanks a lot.


      In case you still needed it

  8. This comment has been removed by the author.

  9. Thank you so much for sharing all this wonderful information !!!! It is so appreciated!! You have good humor in your blogs. So much helpful and easy to read!
    Google Cloud Platform course in pune

  10. Really very happy to say, your post is very interesting to read. I never stop myself to say something about it.You’re doing a great job. Keep it up. Kindly Visit our Website:- AOL Desktop Gold Not Responding

  11. Java is the first choice of many software developers for writing applications for the enterprise. The application of java is essential for enterprise application development. Java Enterprise Edition (Java EE) is a very popular platform that provides API and runtime environment for scripting. It also includes network applications and web services. JavaEE is also considered the backbone for various banking applications that have Java running on the UI to back server end