Metareflect is a lightweight reflection system for C++, based on LLVM and Clangs libtooling.
- Lookup members of reflected classes: Once reflection data has been generated, you can query the members of the class, inspect their types, size and where they're located in the class, as well as qualifiers and storage class.
- Control over which classes are reflected: By annotating your class with
CLASS
, and each member which you want to be reflected withPROPERTY
, you have fine-grained control over what you want to reflect, and what you want to keep completely private. - Extensible: It's easy to add new flags, which can be queried at runtime. This allows for easily adding any feature you like. Currently used to provide serialization from and to a byte stream for annotated classes.
To get started download the latest release here.
Metareflect requires the runtime, it's consisting of the interface and a couple
helper macros for annotating your classes. To include it into your project,
simply copy over the /metareflect
folder into your project.
Unfortunately, the runtime is not fully header-only and requires you to compile
the files ending in .cxx
.
Every class which you want to be reflected needs to be annotated.
point.hxx:
#include <metareflect/metareflect.hxx>
CLASS() Point
{
public:
PROPERTY()
int x;
PROPERTY()
int y;
PROPERTY()
int z;
FUNCTION()
size_t Hash() const
{
return x ^ y ^ z;
}
};
point.cxx:
#include "point.hxx"
#include "point.generated.hxx"
/* rest of the code */
For a full example take a look at our /example
.
To provide the reflection data, the metareflect tool generates a header-file which contains the reflection data in a format consumable by the runtime. The generated file needs to be in your include path and included into the implementation file of the reflected class.
Since metareflect is based on libtooling, all the flags common to libtooling
tools apply to it to. That means if you have a build system which can generate
a compile_commands.json
(for example: ninja or CMake) you can simply provide
it the path to your compilation database and metareflect will pick the correct
flags automatically.
For further information, consult the LLVM documentation for libtooling.
To generate a compilation database using CMake, pass it
-DCMAKE_EXPORT_COMPILE_COMMANDS=ON
while generating your build files.
Any contribution is welcome. Take a look at the open tickets to get started with the development of metareflect.
- A clone of the LLVM 6.0.0 source code, including clang and clang-extra-tools. See LLVM_SETUP how to setup LLVM for development.
- A clone of the metareflect repository
- A beefy computer, otherwise compiling LLVM will take some time
Metareflect consists of two parts, the runtime and the libtooling tool.
The runtime lives in the metareflect/
folder, while the tooling/
folder
contains the source code for the tool.
Once you've cloned the LLVM repo (by following the guide at LLVM_SETUP),
navigate to path/to/llvm/tools/clang/tools/extra/metareflect/metareflect
.
The directory contains the source code for metareflect and anything you need
to get started developing metareflect.
The following resources give an insight into how to develop an libtooling application:
To contribute your changes back, please open a pull request! We welcome any contribution.
Parts of the design and how the tool works has been based upon prior research
done in Unreal Engine 4s UHT (Unreal Header Tool) and Qts moc
.
If you've used either of the two, you might spot the similarities.
Copyright (c) 2018 Arvid Gerstmann.