Considering the following example of using LuaBridge to pass objects to a lua script:
class Test {
public:
double d;
Test(double t): d(t) {};
};
Test t1(10);
auto lua_state = luaL_newstate();
luaL_openlibs(lua_state);
luabridge::getGlobalNamespace(lua_state)
.beginNamespace("Core")
.beginClass<Test>("Test")
.addConstructor<void(*)(double)>()
.addData("d", &Test::d)
.endClass()
.addVariable("t1", &t1)
.endNamespace()
;
Apparently the passed variable can not be modified by lua:
Core.t1.d = 12
print(Core.t1.d)
The above code prints 10
.
Also in C++ after processing the script the variable still holds its initial value.
luaL_dofile(lua_state, ".../test.lua");
std::cout << t1.d << std::endl; // 10
However simple scalar data types (like float) that aren't object members can be modified. Am I missing something?
Edit:
It is apparently modifiable if I replace Test t1(10);
with Test *t1 = new Test(10);
or even Test t(10);
Test *t1 = &t;
and still pass &t1
to addVariable. I do not really understand that behavior, may someone explain? (address of address?)
LuaBridge makes a copy of your argument. If you define a copy constructor yourself and put a breakpoint there, you'll jump from
/**
Push T via copy construction from U.
*/
template <class U>
static inline void push (lua_State* const L, U const& u)
This is just a contract of passing parameters in LuaBridge. I guess, it is so, since the datum has the C++ managed lifetime, and this is a safety feature. You can rewrite your datum with a new value, and you'll get what you want:
Core.t1.d = 12
t2 = Core.Test(10)
t2.d=12
print(Core.t1.d)
print(t2.d)
Core.t1 = t2
print(Core.t1.d)
outputting
10
12
12
and your C++ - side t1 will have the d value of 12