Re: passing pointers using LuaBridge

[Jeremy Ong <jeremycong@...>]

I just noticed you use MSVC9.... carry on then haha. That compiler
isn't C++11'ified enough.

On Mon, Apr 7, 2014 at 1:42 AM, Jeremy Ong <jeremycong@gmail.com> wrote:
> Out of curiosity, have you tried Selene? You can register classes and
> pass pointers to those classes. See the test case at:
>
> https://github.com/jeremyong/Selene/blob/master/test/class_tests.h#L103
>
> The definitions of the Bar and Zoo classes of that test case are at
> the top of that file. Aside from those definitions, the code in that
> test case is self-contained.
>
> On Mon, Apr 7, 2014 at 1:38 AM, Andrew Starks <andrew.starks@trms.com> wrote:
>>
>>
>> On Monday, April 7, 2014, Moose <moose@posteo.de> wrote:
>>>
>>> Hi all,
>>>
>>> I am making moderate progress integrating Lua into my C++ environment.
>>> After trying many different ways of wrapping C or C++ code I ended up with
>>> two candidates that seemed to suit my needs best. LuaBridge
>>> (https://github.com/vinniefalco/LuaBridge) and LuaWrapper
>>> (https://bitbucket.org/alexames/luawrapper).
>>>
>>> I was able to create C++ wrappers for existing classes with both, and,
>>> inside my scripts, to create objects of those types and call methods upon
>>> them. So far so good. However, with both implementations I have failed to
>>> pass existing objects with C++ lifetime into the script.
>>>
>>> With LuaBridge for example I do this (simplified):
>>>
>>> int luaopen_mylib_1(lua_State *L) {
>>>
>>>         using namespace luabridge;
>>>         getGlobalNamespace(L)
>>>                 .beginNamespace("mylib")
>>>                         .addFunction ("testfree", testfree)
>>>                         .beginClass<MyClass>("MyClass")
>>>                                 .addFunction("foo",   &MyClass::foo   )
>>>                         .endClass()
>>>                 .endNamespace();
>>>
>>>         // quite like this interface btw.
>>>
>>>         return 0;
>>> }
>>>
>>> ... then ...
>>>
>>>    using namespace luabridge;
>>>    lua_State *L = luaL_newstate();
>>>    luaL_openlibs(L);
>>>
>>>    // and register the functions within that context
>>>    luaopen_mylib_1(L);
>>>
>>>    // create MyClass instance
>>>    MyClass *p = new MyClass();
>>>    setGlobal(L, p, "mc");     // a LuaBridge Method that pushes the
>>> pointer as
>>>                                          // global userdata on the stack
>>>
>>>    if (luaL_dofile(L, "luawrapper_reuse.lua")) {
>>>                 std::cerr << "Lua program returned an error: \n";
>>>                 std::cerr << lua_tostring(L, -1);
>>>                 std::cerr << std::endl;
>>>                 lua_close(L);
>>>    }
>>>
>>>    delete p;
>>>
>>>
>>>
>>> All very straight forward.
>>> Then in my script...
>>>
>>> [[
>>> print("Executing LuaBridge test")
>>> mylib.testfree()
>>> mc:foo()
>>> ]]
>>>
>>> The method testfree() is being called all right. At least I see the
>>> output. But then, when I try to access the instance "mc", I get:
>>>
>>> luawrapper_reuse.lua:6: attempt to index global 'mc' (a userdata value)
>>>
>>>
>>> Despite the fact that I do everything by the book, exactly the way it is
>>> described in LuaBridges manual. Can anyone tell me what I do wrong? How can
>>> I access the object?
>>> It seems like Lua does know it. A pointer by the name 'mc' has been pushed
>>> as userdata into the context, which is how I understand this is supposed to
>>> work. But why can I not de-reference it and call a method?
>>> So close to the goal and yet all my efforts fail on this....
>>>
>>> With that Test I use LuaJit 2.0 on Windows64 (MSVC9).
>>>
>>>
>>> Help is much appreciated.
>>>
>>> Thanks,
>>> Moose.
>>>
>>>
>>>
>>
>> Smart guy question: have you tried "Lua the C API?"
>>
>> But that's an question from someone who only trusts his own abstractions...
>> Nothing to see here...