I'm writing a simple, lightweight engine in D. For the input calls I use GLFW3. The library in question uses callbacks to send input events to the program.
What I would like is to use a method from a class as the callback function, rather than a function. This is proving difficult (just as it is in C++). I believe there is an elegant way to do it, but this is how I got it right now.
public void initialise(string logPath) {
[...]
m_Window = new RenderWindow();
m_Window.create();
// Lets set up the input loop.
GLFWkeyfun keyCB = function(GLFWwindow* win, int key, int scancode, int action, int mods) {
printf("Got key event: %d:%d:%d:%d\n");
RenderWindow rw = Root().getRenderWindow();
switch (key) {
case KeyboardKeyID.Q:
glfwSetWindowShouldClose(win, true);
break;
case KeyboardKeyID.H:
if (rw.hidden) {
rw.show();
} else {
rw.hide();
}
break;
default:
break;
}
};
glfwSetKeyCallback(m_Window.window, keyCB);
}
Here is the definition of the callback setting function and type:
extern (C) {
alias GLFWkeyfun = void function(GLFWwindow*, int, int, int, int);
GLFWkeyfun glfwSetKeyCallback(GLFWwindow*, GLFWkeyfun);
}
What I would like to do instead, is create a method that is part of the class. Is there any way to do this?
A solution I tried was a static method wrapped around in extern (C), this worked for calling it, but then I could (obviously) not access this or any other methods, which defeats the point of the exercise.
Thanks in advance.
The way I'd do it is to have a static map of the pointers to the class, so like:
static YourWindowClass[GLFWwindow*] mappings;
Then, in the constructor, once you get a GLFWwindow pointer, add it right in:
mappings[m_Window.window] = this;
Now, make the static extern(C) function to use as the callback. When it gets a pointer from C, look up your class reference in that mappings array and then go ahead and call the member function through that, forwarding the arguments.
So a bit of an extra step, but since it doesn't look like the callback lets you pass user-defined data to it (BTW, attention all lib writers: user-defined void* to the callbacks is sooooo useful, you should do it whenever possible!), but since it doesn't do that the associative array is the next best thing.
Well, I have figured it out my own. The solution I went with was a Singleton class InputManager. Instances of RenderWindow attach themselves to it with the following function. The InputManager then creates an anonymous function() for the RenderWindow that receives events, which then calls a function that handles the actual event.
The idea is then that listeners attach themselves to the InputManager and receive keyboard events for the RenderWindow they requested.
class InputManager {
private static InputManager m_Instance;
private RenderWindow[] m_Watched;
private KeyboardListener[][RenderWindow] m_KeyListeners;
public void recvKeyEvent(GLFWwindow* w, int k, int c, int a, int m) {
writeln("Received key: ", k);
}
public void watch(RenderWindow win) {
if (!isWatched(win)) {
// Relay the key callbacks onto the InputManager.
GLFWkeyfun keyCB = function(GLFWwindow* w, int k, int c, int a, int m) {
InputManager().recvKeyEvent(w, k, c, a, m);
};
glfwSetKeyCallback(win.window, keyCB);
}
}
private bool isWatched(RenderWindow win) {
foreach(RenderWindow w; m_Watched) {
if (win == w) {
return true;
}
}
return false;
}
public static InputManager opCall() {
if (m_Instance is null) {
m_Instance = new InputManager();
}
return m_Instance;
}
private this() {
// nothing
}
}
Works like a charm, now to figure out how to properly attach listeners elegantly.
For those curious, the full source code with how this is set up can be found at https://github.com/Adel92/Mage2D. I hope it helps someone else in a similar position with callbacks.
Related
I have a struct A that inherits from other classes (which I'm not allowed to change). Inside A and it's methods I can call inherited methods (lets say A_method(int i), for example) without problem but when I tried to write a nested struct (lets say In) and call A_method(int i) and there is were I'm stuck.
The initial code looks like this, and I can't change it, is some kind of college assigment.
#include "Player.hh"
struct A : public Player {
static Player* factory () {
return new A;
}
virtual void play () {
}
};
RegisterPlayer(PLAYER_NAME);
Then I tried this:
#include "Player.hh"
struct A : public Player {
static Player* factory () {
return new A;
}
//My code
struct In {
int x;
void do_smthing() {
A_method(x);
}
}
virtual void play () {
}
};
RegisterPlayer(PLAYER_NAME);
Ok, from a beginning I knew I could't do this, for In to see it's parent class it should have a pointer to it but In is a often instantiated object in my code and I wanted to avoid passing this constantly to a constructor so I tried this aproach:
#include "Player.hh"
struct A : public Player {
static Player* factory () {
return new A;
}
//My code
static struct Aux
A* ptr;
Aux(A* _p) { ptr = _p; }
} aux;
struct In {
int x;
void do_smthing() {
aux.ptr->A_method(x);
}
}
virtual void play () {
//the idea is to call do_smthing() here.
}
};
RegisterPlayer(PLAYER_NAME);
What I want to avoid (if possible) is something like this:
struct In {
int x;
A* ptr;
In (A* _p) : ptr(_p) {}
void do_smthing() {
ptr->A_method(x);
}
}
The main reason for this: I have more struct definitions and they they are instantiated multiple times through the rest of the (omitted) code, and I don't like the idea of seeing In(this) so many times.
I don't know if I'm completly missing something or what I want to do it's just not possible... Please ask for clarifications if necessary.
(Also, performance is kind of critical, my code will be tested with limited CPU time so I kinda have to avoid expensive approachs if possible. Using C++11)
There is no way you can skip passing the this pointer. Instead, you could create a helper function in A:
template <typename InnerType, typename ...Params>
InnerType makeInner(Params&&... params)
{
return InnerType(this, std::forward<Params>(params)...);
}
Then you can use
auto * a = A::factory();
auto inner = a->makeInner<A::In>();
I have some suggestions which are not directly related to you question but may help:
A::facotry() returns a std::unique_ptr<A> instead of raw pointer
Try to describe what problem you are trying to solve. I have a strong feeling that there can be a better design other than creating many nested structs.
I don't see passing a this pointer could have any impact on the performance. The more important thing is to identify the path that is latency-sensitive and move expensive operations out of those paths.
The crux of the issue is I want to create a vector of base pointers to reference children objects. However I'm having issues accessing the methods of the children. I've seen examples of downcasting online but I don't feel it's the best thing for me since I want to keep my code generic. Please look below for a sample of what I'm trying to accomplish.
class Base
{
public:
stuffx;
private:
stuffy;
}
template<typename U>
class Child : public Base
{
public:
Child(
std::function<U()> getterFunc,
std::function<void(U)> setterFunc
):
mgetter(getterFunc),
msetter(setterFunc)
{
}
U getFunction() const {return m_getter();}
void setFunction(U input) const {return m_setter(input);}
private:
observableValues() {}
std::function<U()> m_getter;
std::function<void(U)> m_setter;
}
int main()
{
std::vector<std::shared_ptr<Base>> Dummy = {std::make_shared<Child<int>> (std::bind(..), std::bind(...)),
std::make_shared<Child<string>> (std::bind(..), std::bind(...)) };
Dummy.at(0)->getGFunction(); // this throws an error as out of scope.
(dynamic_cast<Child<int>>(Dummy.at(0))->getGFunction(); // this is ok
}
In this example above my vector is of size 2 which is manageable but my goal is to serialize c++ classes to a psql server and may have to handle vectors of size 30+. My next question is is there a way to automate this in a for loop taking into the account the type deduction that may need to be performed for typename U.
int main()
{
std::vector<std::shared_ptr<Base>> Dummy = {std::make_shared<Child<int>> (std::bind(..), std::bind(...)),
std::make_shared<Child<string>> (std::bind(..), std::bind(...)) };
std::vector<std::shared_ptr<Base>>::const_iterator it_base = Dummy.begin();
for (; it_base != Dummy.end(); ++it_base)
{
//insert method here for downcasting
}
}
There is a custom defined map, with an element std::function()>.
The lambda code is working, but I don't know how to expand it to a normal formation. The code is following.
class TestA{
public:
TestA() {}
~TestA() {}
TestA(const TestA &) {}
static void print()
{
cout << __FUNCTION__ << endl;
return;
}
};
void testComplexMap1()
{
typedef map<string, std::function<std::unique_ptr<TestA>()>> TempMap;
TempMap m;
// the lambda format code, it works
//m.insert({ "TestA", []() {return std::unique_ptr<TestA>(new TestA());}});
// I want to expand it, but failed.
TestA *t = new TestA();
//function<unique_ptr<TestA>()> fp(unique_ptr<TestA>(t));
function<unique_ptr<TestA>()> fp(unique_ptr<TestA>(t)()); //warning here
//m.emplace("TestA", fp); // compile error here
}
Any help will be greatly appreciated.
fp is not initialized with a function so compilation fails.
You can expand it like this:
TestA *t = new TestA();
std::unique_ptr<TestA> UT(t);
auto func = [&]() { return move(UT);};
std::function<std::unique_ptr<TestA>()> fp(func);
m.emplace("TestA", fp);
See DEMO.
In C++ everything that looks like it could be a declaration is treated as such.
This means the line
function<unique_ptr<TestA>()> fp(unique_ptr<TestA>(t)());
is interpreted as:
fp is the declaration of a function returning an std::function<unique_ptr<TestA>()> and expecting a parameter called t which is a function pointer to a function returning a std::unique_ptr<TestA> and getting no parameter. (Which is not what you intended.)
This also means that the t in this line is not the same t as in the previous line.
You have to pass fp something that is actually callable like this:
std::unique_ptr<TestA> f() {
return std::make_unique<TestA>();
}
void testComplexMap1() {
// ...
function<unique_ptr<TestA>()> fp(f);
m.emplace("TestA1", fp);
}
If you want to add a function to the map that wraps an existing pointer into a unique_ptr you would need either a functor:
class Functor {
public:
Functor(TestA * a) : m_a(a) {}
~Functor() { delete m_a; }
std::unique_ptr<TestA> operator()(){
auto x = std::unique_ptr<TestA>(m_a);
m_a = nullptr;
return std::move(x);
}
private:
TestA * m_a;
};
void testComplexMap1() {
//...
TestA * t = new TestA();
m.emplace("TestA", Functor(t));
}
Or a lambda with capture:
void testComplexMap1() {
//...
TestA * t = new TestA();
m.emplace("TestA", [t](){ return std::unique_ptr<TestA>(t); });
}
The lamda is translated more or less to something like the Functor class. However in each case you have to be really careful: The functions in the map that encapsulate an existing pointer into a std::unique_ptr can and should only be called once.
If you don't call them, memory allocated for t won't be freed. If you call them more than once you get either a std::unique_ptr to nullptr (in my Functor class variant) or a more than one std::unique_ptr tries to manage the same memory region (in the lambda with capture variant), which will crash as soon as the second std::unique_ptr is deleted.
In short: I would advice against writing code like this and only put functions in the map that are callable multiple times.
I recently finished a 6-month internship at a company that uses C# for the most part of their programming. During this time I first used and got accustomed to the C# way of doing events. Like shown below:
acc.AccountBalanceLow += new AccountBalanceDelegate(atm.AccountToLow);
acc.AccountBalanceLow +=new AccountBalanceDelegate(atm.AccountToLowAgain);
Does D support such constructs? I'd imagine one could be created by the user by using operator overloading, but I'm not entirely sure. If it's not possible what would then be a common excepted way of doing it then?
The equivalent construct in D is to use Signals and Slots. This is a different means of implementing the Observer Pattern, which is effectively what a C# event does.
D (and C++) use an analogous pattern called signals and slots.
If you're feeling the need to use the C# style-events instead of signals and slots, they're extremely simple to implement:
module fluidity.core.event;
class Event {
alias void delegate(EventArgs) handler_t;
handler_t[] handlers;
Object owner;
this() {}
this(Object o) { owner = o; }
void attach(handler_t handler) {
if (handler)
handlers ~= handler;
}
void detach(handler_t handler) {
int i = -1;
foreach (j, h; handlers)
{
if (h is handler)
{
i = j;
break;
}
}
if (i > -1)
handlers = handlers[0..i] ~ handlers[i+1..$];
}
void raise() { raise(new EventArgs(owner)); }
void raise(EventArgs e) {
// call all handlers
foreach (handler; handlers)
{
if (handler)
handler(e);
}
}
void opAddAssign(handler_t handler) {
attach(handler);
}
void opSubAssign(handler_t handler) {
detach(handler);
}
}
class EventArgs {
Object source;
bool handled;
void handle() { handled = true; }
this() {}
this(Object s) {
source = s;
}
}
Here is an example of c# style events using signals, slots, and a templates:
events.d:
import std.signals;
class Event(T...){
mixin Signal!(T);
void broadcast(T args){
emit(args);
}
void opAddAssign(slot_t slot){
connect(slot);
}
void opSubAssign(slot_t slot) {
disconnect(slot);
}
}
declaration:
public Event!(int) onSomeEventOfInt;
public Event!(string, int) onSomeEventOfStringAndInt;
instantiation:
this.onSomeEventOfInt = new Event!(int)();
this.onSomeEventOfStringAndInt = new Event!(string, int)();
fire event:
int i = 4;
string str = "hello";
this.onSomeEventOfInt.broadcast(i);
this.onSomeEventOfStringAndInt.broadcast(str, 4);
observer registration:
obj1.onSomeEventOfInt += &handleEventOfInt
obj1.onSomeEventOfStringAndInt += &handleEventOfStringAndInt
void handleEventOfInt(int g)
{ /*do something */ }
void handleEventOfStringAndInt(string str, int g)
{ /*do something */ }
Check out DFL's event system. It works EXACTLY the same way as C# .NET.
DFL Event Example
Download DFL, grab the events module and use it the way you like. I modified it to use variadic template arguments. This gives maximum flexibility.
http://www.dprogramming.com/dfl098.zip
I was wondering what's the proper way of raising events from C++/CLI. In C# one should first make a copy of the handler, check if it's not null, and then call it. Is there a similar practice for C++/CLI?
This isn't the whole story! You don't usually have to worry about null event handlers in C++/CLI. The code for these checks is generated for you. Consider the following trivial C++/CLI class.
public ref class MyClass
{
public:
event System::EventHandler ^ MyEvent;
};
If you compile this class, and disassemble it using Reflector, you get the following c# code.
public class MyClass
{
// Fields
private EventHandler <backing_store>MyEvent;
// Events
public event EventHandler MyEvent
{
[MethodImpl(MethodImplOptions.Synchronized)] add
{
this.<backing_store>MyEvent = (EventHandler) Delegate.Combine(this.<backing_store>MyEvent, value);
}
[MethodImpl(MethodImplOptions.Synchronized)] remove
{
this.<backing_store>MyEvent = (EventHandler) Delegate.Remove(this.<backing_store>MyEvent, value);
}
raise
{
EventHandler <tmp> = null;
<tmp> = this.<backing_store>MyEvent;
if (<tmp> != null)
{
<tmp>(value0, value1);
}
}
}
}
The usual checks are being done in the raise method. Unless you really want custom behavior, you should feel comfortable declaring your event as in the above class, and raising it without fear of a null handler.
C++/CLI allows you to override raise in custom event handlers so you don't have to test for null or copy when raising the event. Of course, inside your custom raise you still have to do this.
Example, adapted from the MSDN for correctness:
public delegate void f(int);
public ref struct E {
f ^ _E;
public:
void handler(int i) {
System::Console::WriteLine(i);
}
E() {
_E = nullptr;
}
event f^ Event {
void add(f ^ d) {
_E += d;
}
void remove(f ^ d) {
_E -= d;
}
void raise(int i) {
f^ tmp = _E;
if (tmp) {
tmp->Invoke(i);
}
}
}
static void Go() {
E^ pE = gcnew E;
pE->Event += gcnew f(pE, &E::handler);
pE->Event(17);
}
};
int main() {
E::Go();
}
If your issue is that raise isn't private, then explicitly implement it like the docs say:
http://msdn.microsoft.com/en-us/library/5f3csfsa.aspx
In summary:
If you just use the event keyword, you create a "trivial" event. The compiler generates add/remove/raise and the delegate member for you. The generated raise function (as the docs say) checks for nullptr. Trivial events are documented here:
http://msdn.microsoft.com/en-us/library/4b612y2s.aspx
If you want "more control", for example to make raise private, then you have to explicitly implement the members as shown in the link. You must explicitly declare a data member for the delegate type. Then you use the event keyword to declare the event-related members, as in the Microsoft example:
// event keyword introduces the scope wherein I'm defining the required methods
// "f" is my delegate type
// "Event" is the unrealistic name of the event itself
event f^ Event
{
// add is public (because the event block is public)
// "_E" is the private delegate data member of type "f"
void add(f ^ d) { _E += d; }
// making remove private
private:
void remove(f ^ d) { _E -= d; }
// making raise protected
protected:
void raise(int i)
{
// check for nullptr
if (_E)
{
_E->Invoke(i);
}
}
}// end event block
Wordy, but there it is.
-reilly.