TechOverflow Logo
TechOverflow

How to export C++ struct fields to JSON using clang and boost::json

In our previous post How to parse C++ struct definition from source code using clang, we learned how to parse C/C++ code and extract information about struct definitions using the clang library.

In said post, the information was output in a freeform, human-readable format, however:

Struct: myParameters
  Field: a (Type: double)
  Field: b (Type: double)

In this post, we will extend the previous example to export the struct fields to JSON format using the boost::json library.

Full source code

#include <clang-c/Index.h>
#include <iostream>
#include <vector>
#include <string>
#include <cstring>
#include <fstream>
#include <filesystem>
#include <boost/json.hpp>

boost::json::value extractStructFields(CXCursor cursor) {
    CXCursorKind kind = clang_getCursorKind(cursor);
    if (kind == CXCursor_StructDecl) {
        CXString structName = clang_getCursorSpelling(cursor);
        std::string structNameStr = clang_getCString(structName);
        clang_disposeString(structName);
        
        boost::json::object structObj;
        structObj["name"] = structNameStr;
        boost::json::array fieldsArray;
        
        struct FieldCollector {
            boost::json::array& fields;
            int currentIndex;
        };
        
        FieldCollector collector = {fieldsArray, 0};
        
        clang_visitChildren(cursor, [](CXCursor c, CXCursor parent, CXClientData client_data) {
            auto* collector = static_cast<FieldCollector*>(client_data);
            CXCursorKind kind = clang_getCursorKind(c);
            
            if (kind == CXCursor_FieldDecl) {
                boost::json::object fieldObj;
                
                CXString fieldName = clang_getCursorSpelling(c);
                CXType fieldType = clang_getCursorType(c);
                CXString typeName = clang_getTypeSpelling(fieldType);
                
                fieldObj["index"] = collector->currentIndex++;
                fieldObj["name"] = clang_getCString(fieldName);
                fieldObj["type"] = clang_getCString(typeName);
                
                collector->fields.push_back(fieldObj);
                
                clang_disposeString(fieldName);
                clang_disposeString(typeName);
            }
            return CXChildVisit_Continue;
        }, &collector);
        
        structObj["fields"] = fieldsArray;
        return structObj;
    }
    return boost::json::value(nullptr);
}

int main(int argc, char** argv) {
    CXIndex index = clang_createIndex(0, 0);

    // Read code from argv[1]
    if (argc < 2) {
        std::cerr << "Usage: " << argv[0] << " <source file>" << std::endl;
        return 1;
    }
    const char* codeFilename = argv[1];
    if(!std::filesystem::exists(codeFilename)) {
        std::cerr << "File does not exist: " << codeFilename << std::endl;
        return 1;
    }
    std::ifstream codeFile(codeFilename);
    if (!codeFile) {
        std::cerr << "Error opening file: " << codeFilename << std::endl;
        return 1;
    }
    std::string code((std::istreambuf_iterator<char>(codeFile)), std::istreambuf_iterator<char>());
    codeFile.close();
    if (code.empty()) {
        std::cerr << "File is empty: " << codeFilename << std::endl;
        return 1;
    }

    CXUnsavedFile unsavedFile = {"test.cpp", code.c_str(), code.size()};
    CXTranslationUnit unit = clang_parseTranslationUnit(index, "test.cpp", nullptr, 0, &unsavedFile, 1, CXTranslationUnit_None);
    if (unit == nullptr) {
        std::cerr << "Failed to parse translation unit." << std::endl;
        return 1;
    }

    CXCursor cursor = clang_getTranslationUnitCursor(unit);
    boost::json::array structsArray;
    
    clang_visitChildren(cursor, [](CXCursor c, CXCursor parent, CXClientData client_data) {
        auto* structs = static_cast<boost::json::array*>(client_data);
        boost::json::value structJson = extractStructFields(c);
        
        if (!structJson.is_null()) {
            structs->push_back(structJson);
        }
        return CXChildVisit_Continue;
    }, &structsArray);

    boost::json::object result;
    result["structs"] = structsArray;
    std::cout << boost::json::serialize(result) << std::endl;

    clang_disposeTranslationUnit(unit);
    clang_disposeIndex(index);
    return 0;
}

How to compile

In this case, we used cmake to build the project. The CMakeLists.txt file looks like this:

cmake_minimum_required(VERSION 3.10)
project(parse_structs)

# C++17 Standard festlegen
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

# LLVM/Clang-Komponenten finden
find_package(LLVM 19.1.1 REQUIRED CONFIG)
find_package(Clang REQUIRED CONFIG)
find_package(Boost REQUIRED COMPONENTS json)

# Include-Verzeichnisse hinzufügen
include_directories(${LLVM_INCLUDE_DIRS})
include_directories(${CLANG_INCLUDE_DIRS})

# Executable erstellen
add_executable(parse_structs parse_structs.cpp)

# LLVM-Definitionen hinzufügen
target_compile_definitions(parse_structs PUBLIC ${LLVM_DEFINITIONS})

# Gegen libclang linken
target_link_libraries(parse_structs PUBLIC libclang ${Boost_LIBRARIES})

How to run

./parse_structs test.cpp | jq

| jq is used to pretty-print the JSON output. You don’t need to use it if you don’t want to pretty-print the output.

with test.cpp containing, for example, the following code:

struct myParameters {
    double a;
    double b;
};

Example output

{
  "structs": [
    {
      "name": "myParameters",
      "fields": [
        {
          "index": 0,
          "name": "a",
          "type": "double"
        },
        {
          "index": 1,
          "name": "b",
          "type": "double"
        }
      ]
    }
  ]
}