physx-c is a handcrafted C wrapper over the PhysX 4.1 SDK.
physx-c aims to faithfully reproduce PhysX by matching terminology for classes/structs, functions, variables, and even function arguments.
physx-c is not a higher level abstraction over PhysX nor does it try to simplify or improve any choices made by PhysX. Instead the aim is expose the SDK through a C API with minimal overhead, in a way that enables people familiar with PhysX to get started without having to learn anything new, and to enable newcomers to use existing PhysX resources.
With that being said, a few 'utility' functions like NewCPxSomeClass are provided, but their usage is completely optional.
PhysX uses the naming scheme PxSomeName. physx-c simply prefixes a 'C' to make the equivalent C struct CPxSomeName.
For example, PhysX classes are usually in the format PxSomeClass. physx-c wraps these using C structs in the format CPxSomeClass with the
C struct usually holding a void* to the C++ class. This general rule is only broken in cases where the C++ equivalent is a simple
value type, like vectors or quaternions. In these cases the C struct holds the actual data (e.g. x,y,z fields) and is converted into
the proper C++ class by physx-c transparently where needed. This helps avoid unnecessary heap allocations.
Similarly, PhysX macros and typedefs can have physx-c parallels, like PxU32 being matched by CPxU32.
Member functions of classes (aka methods) are wrapped in the format CPxSomeClass_funcName, where 'funcName' matches the name
used by PhysX. A pointer to the CPxSomeClass struct is always taken as the first argument, with the rest of the arguments matching PhysX.
Member variables are accessed through getters and setters (where available) in the format CPxSomeClass_get_myVar/CPxSomeClass_set_myVar.
Similar to methods, a pointer to the CPxSomeClass struct is always taken as the first argument.
physx-c uses physx::PxDefaultAllocator for all allocations through the CPxAlloc/CPxDealloc macros defined in PxDefaultAllocator.h. If you would like to use a custom allocator then you can change the header to add your own implementation.
By default, physx-c builds into a single physx-c.dll file which then loads the needed PhysX DLLs at startup.
To use it simply add the include folder to your includes and make sure physx-c.dll and the normal PhysX DLLs (e.g. PhysX_64.dll) are
in the same folder (usually in the same folder as your executable).
You can get some prebuilt PhysX DLLs from the PhysX repo, or you can build the source from the original PhysX repo, or from the releases on this repo.
To see a real-world usage of physx-c, you can checkout physx-go for an example of how to use physx-c to provide a PhysX wrapper for a different programming language (Go in this case).
The absolute core features of PhysX have all been wrapped, which includes scenes, collision shapes, raycasts, rigid bodies, visual debugger, and simulations, among others.
Some of the things that have not been wrapped include:
- Joints (planned)
- Articulations
- Vehicles
- Character controller
- Serialization
- Simulation Statistics
PRs are welcome!