Skip to content

Latest commit

 

History

History
98 lines (59 loc) · 3.51 KB

installation.md

File metadata and controls

98 lines (59 loc) · 3.51 KB

Installation

The project consits of two main parts:

The Easy Reflection C++ Generator utilizes Clang libraries to analyze C++ source code. By default, these libraries are statically linked to facilitate distribution and provide precompiled binary releases.

Before getting started, you should decide whether to use the precompiled binary release or build the generator manually. If you choose the manual option, configure and build the generator. CMake will handle everything for you.

Once you have the er_gen binary, you can start analyzing your project and generating code. To do this, ensure that your project produces a compile_commands.json file. If you're using CMake, you can easily enable this by adding the following line to your CMakeLists.txt file:

set(CMAKE_EXPORT_COMPILE_COMMANDS ON)

Troubleshooting

If you encounter errors like stddef.h or stdarg.h not being found during the generation process, add the following line to your CMakeLists.txt file:

set(CMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES ${CMAKE_CXX_IMPLICIT_INCLUDE_DIRECTORIES})

Windows

Visual Studio CMake generators do not create compile_commands.json. However, you can use Ninja instead.

To build er_gen and use reflection on Windows, make sure to install the following dependencies and add them to your PATH:

  • Git
  • CMake
  • Ninja

If you choose the MSVC compiler, additionally install:

  • C++ ATL
  • Windows SDK

Docker

This repository includes a Dockerfile that sets up an Ubuntu 22.04 environment, builds the generator, and runs tests on startup.

Further steps

Another submodule is vcpkg which manages most of the dependencies, all of them will be installed by CMake automatically.

After installing all the dependencies, you need to decide whether you want to use simdjson for parsing or not. While simdjson provides the fastest performance on supported hardware, rapidjson might be faster in virtual environments. The solution includes a native parser that is available regardless. It offers more flexibility when it comes to map parsing but not as fast.

If, for some reason, you want to exclude simdjson from the dependencies, you can use the following CMake option:

-DUSE_SIMD_JSON=OFF

LLVM & Clang dynamic linking

It is not supported officialy but nevertheless it exists.

Replace lines in the end of CMakeLists.txt

find_package(LLVM REQUIRED)

include_directories(${LLVM_INCLUDE_DIR})
link_directories(${LLVM_LIBRARY_DIR})

target_link_libraries(${PROJECT_NAME} PRIVATE LLVM clang-cpp)

Linux

You can install libraries from package manager of your distro such as Ubuntu or Arch Linux or build them from source:

cd llvm-project/build

cmake -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_PROJECTS=clang -DLLVM_ENABLE_RTTI=ON -DLLVM_LINK_LLVM_DYLIB=ON -DCLANG_LINK_CLANG_DYLIB=ON ../llvm

Apple

Apple's macOS doesn't have standard Clang libs by default, but happily you can use brew.

brew install llvm

Don't forget to specify the path to LLVM.

-DLLVM_DIR=/opt/homebrew/opt/llvm/lib/cmake/llvm

Windows

Regrettably, MSVC does not support for this feature. However, there is some dark magic with MinGW and Clang you can delve into. Unfortunately, I cannot provide assistance with that particular matter.