引言:如果你想深入了解cocos2d-x的整个框架和运行流程,如果你想知道整个启动过程的细节,如果你想知道自己写的代码是在什么时候和在哪里被调用的,下面可以为你解答其中奥秘。
对象:适合刚刚入门了cocos2d-x的初学者,编写并运行过简单的demo,并且想仔细探究其中的原理机制。
1、程序入口main.cpp
- int APIENTRY _tWinMain(HINSTANCE hInstance,
- HINSTANCE hPrevInstance,
- LPTSTR lpCmdLine,
- int nCmdShow)
- {
- ...
- // create the application instance
- AppDelegate app;//看这里
- ...
- return CCApplication::sharedApplication()->run();//看这里
- }
以上两行是关键代码
(1)代码AppDelegate app;调用了AppDelegate的构造函数,而它的父类是CCApplication,所以CCApplication的构造函数也被调用了
- CCApplication::CCApplication()
- : m_hInstance(NULL)
- , m_hAccelTable(NULL)
- {
- m_hInstance = GetModuleHandle(NULL);
- m_nAnimationInterval.QuadPart = 0;
- CC_ASSERT(! sm_pSharedApplication);
- sm_pSharedApplication = this;
- }
这里完成了AppDelegate和CCApplication之间的联系,由于AppDelegate是CCApplication的子类,故CCApplication里面的静态单例指针指的便是AppDelegate的实例。之后调用CCApplication::sharedApplication()的相关操作是基于AppDelegate的实现。
(2)代码CCApplication::sharedApplication()->run();调用了下面的代码
- int CCApplication::run()
- {
- ...
- // Initialize instance and cocos2d.
- if (!applicationDidFinishLaunching())
- {
- return 0;
- }
- ..
- while (1)
- {
- if (! PeekMessage(&msg, NULL, 0, 0, PM_REMOVE))
- {
- // Get current time tick.
- QueryPerformanceCounter(&nNow);
- // If it's the time to draw next frame, draw it, else sleep a while.
- if (nNow.QuadPart - nLast.QuadPart > m_nAnimationInterval.QuadPart)
- {
- nLast.QuadPart = nNow.QuadPart;
- CCDirector::sharedDirector()->mainLoop();//看这里
- }
- else
- {
- Sleep(0);
- }
- continue;
- }
- if (WM_QUIT == msg.message)
- {
- // Quit message loop.
- break;
- }
- // Deal with windows message.
- if (! m_hAccelTable || ! TranslateAccelerator(msg.hwnd, m_hAccelTable, &msg))
- {
- TranslateMessage(&msg);
- DispatchMessage(&msg);
- }
- }
- return (int) msg.wParam;
- }
在这里,run()函数里面是一个while死循环,不过里面有判断条件 if (WM_QUIT == msg.message)可以跳出循环,终止程序。循环里面每隔一段时间m_nAnimationInterval就调用CCDirector::sharedDirector()->mainLoop(),完成游戏的刷新。
(3)下面来看看mainLoop()都干了啥事:
- void CCDisplayLinkDirector::mainLoop(void)
- {
- if (m_bPurgeDirecotorInNextLoop)
- {
- m_bPurgeDirecotorInNextLoop = false;
- purgeDirector();
- }
- else if (! m_bInvalid)
- {
- drawScene();//看这里
- // release the objects
- CCPoolManager::sharedPoolManager()->pop(); //看这里
- }
- }
首先解释一下为什么显示的是CCDisplayLinkDirector::mainLoop,而不是 CCDirector ::mainLoop,答案就在CCDirector.h头文件里面
- class CC_DLL CCDirector : public CCObject, public TypeInfo
- {
- .......
- virtual void mainLoop(void) = 0;//看这里
- .......
- }
- class CCDisplayLinkDirector : public CCDirector
- {
- ..........
- }
在CCDirector类里面,mainLoop是纯虚函数,而CCDisplayLinkDirector是CCDirector的子类。
重新回到mainLoop函数的实现,我们发现drawScene()和CCPoolManager::sharedPoolManager()->pop()是其主要内容,也就是重绘和释放自动内存池管理器里面暂时存储的对象,这个和cocos2d-x的自动内存管理机制有关,所有静态创建对象函数::create()都采用了自动托管机制,比如:
- CCSprite* CCSprite::create(const char *pszFileName)
- {
- CCSprite *pobSprite = new CCSprite();
- if (pobSprite && pobSprite->initWithFile(pszFileName))
- {
- pobSprite->autorelease();//看这里
- return pobSprite;
- }
- CC_SAFE_DELETE(pobSprite);
- return NULL;
- }
其中autorelease()函数实现如下
- CCObject* CCObject::autorelease(void)
- {
- CCPoolManager::sharedPoolManager()->addObject(this);
- return this;
- }
由此可以看到自动内存管理对象在每帧绘制结束后都会被CCPoolManager::sharedPoolManager()->pop()释放,除非我们在create()对象后使用了retain(),retain()也很简单,实现如下
- void CCObject::retain(void)
- {
- CCAssert(m_uReference > 0, "reference count should greater than 0");
- ++m_uReference;//看这里
- }
就是把对象的引用计数-1,这又对应于对象的release()函数
- void CCObject::release(void)
- {
- CCAssert(m_uReference > 0, "reference count should greater than 0");
- --m_uReference;//看这里
- if (m_uReference == 0)//看这里
- {
- delete this;
- }
- }
可以看到,对象只有当其引用计数m_uReference=0时,才会真正被删除。
(4)下面看看drawScene()重绘函数负责什么,看完你就全明白了整个框架了
- // Draw the Scene
- void CCDirector::drawScene(void)
- {
- ..............
- //tick before glClear: issue #533
- if (! m_bPaused)
- {
- m_pScheduler->update(m_fDeltaTime);//看这里
- }
- .....................
- // draw the scene
- if (m_pRunningScene)
- {
- m_pRunningScene->visit();
- }
- .....................
- // swap buffers
- if (m_pobOpenGLView)
- {
- m_pobOpenGLView->swapBuffers();
- }
- ................
- }
大家看到 m_pScheduler->update(m_fDeltaTime)会不会想到我们经常在HelloWorldScene.h写的void update(float dt),两者有什么关系呢?
首先回忆一下,我们在HelloWorldScene.h里面是怎样调用到update函数的
this->schedule(schedule_selector( HelloWorld::update));(这只是调用方式的一种)
看看schedule函数的实现
- void CCNode::schedule(SEL_SCHEDULE selector)
- {
- this->schedule(selector, 0.0f, kCCRepeatForever, 0.0f);
- }
- void CCNode::schedule(SEL_SCHEDULE selector, float interval, unsigned int repeat, float delay)
- {
- CCAssert( selector, "Argument must be non-nil");
- CCAssert( interval >=0, "Argument must be positive");
- m_pScheduler->scheduleSelector(selector, this, interval , repeat, delay, !m_bRunning);//看这里
- }
- class CC_DLL CCNode : public CCObject
- { .........
- CCScheduler *m_pScheduler; //看这里
- .......
- }
再对比一下刚才drawScene()重绘函数
- // Draw the Scene
- void CCDirector::drawScene(void)
- {
- ..............
- //tick before glClear: issue #533
- if (! m_bPaused)
- {
- m_pScheduler->update(m_fDeltaTime);//看这里
- }
- .....................
- }
- class CC_DLL CCDirector : public CCObject, public TypeInfo
- { ...............
- CC_PROPERTY(CCScheduler*, m_pScheduler, Scheduler);//看这里
- ..............
- }
- #define CC_PROPERTY(varType, varName, funName)\//看这里
- protected: varType varName;\
- public: virtual varType get##funName(void);\
- public: virtual void set##funName(varType var);
我们发现CCNode和CCDirector类都申明了CCScheduler *m_pScheduler,都不是同一个滴。
且慢,不一定哦,看看CCNode的构造函数先
- CCNode::CCNode(void)
- : ...........
- {
- CCDirector *director = CCDirector::sharedDirector();
- ...............
- m_pScheduler = director->getScheduler();//看这里
- ................
- }
原来CCNode里面的m_pScheduler引用了CCDirector的m_pScheduler,感受到穷追不舍的魅力了吧。
再回到之前的问题:
CCDirector::drawScene里面的m_pScheduler->update(m_fDeltaTime)和我们经常在HelloWorldScene.h写的void update(float dt),两者有什么关系呢?
由上面代码可见,this->schedule(schedule_selector( HelloWorld::update))实际上会把schedule_selector( HelloWorld::update)保存到CCNode里面的m_pScheduler,也就是CCDirector里面的m_pScheduler,那么CCDirector::drawScene里面的m_pScheduler->update(m_fDeltaTime)里面的代码应该就和HelloWorld::update()有关吧。继续看看呗。
- // main loop
- void CCScheduler::update(float dt)
- {
- ................
- // Iterate over all the custom selectors
- for (tHashTimerEntry *elt = m_pHashForTimers; elt != NULL; )
- {
- m_pCurrentTarget = elt;
- m_bCurrentTargetSalvaged = false;
- if (! m_pCurrentTarget->paused)
- {
- // The 'timers' array may change while inside this loop
- for (elt->timerIndex = 0; elt->timerIndex < elt->timers->num; ++(elt->timerIndex))
- {
- elt->currentTimer = (CCTimer*)(elt->timers->arr[elt->timerIndex]);
- elt->currentTimerSalvaged = false;
- elt->currentTimer->update(dt);//看这里
- ......................
- }
- }
- ..................
- }
- ...............
- }
重点看elt->currentTimer->update(dt);
elt->currentTimer是CCTimer类,看看里面的update函数是神马情况先
- void CCTimer::update(float dt)
- {
- if (m_fElapsed == -1)
- {
- ....................
- }
- else
- {
- if (m_bRunForever && !m_bUseDelay)
- {//standard timer usage
- ...................
- if (m_fElapsed >= m_fInterval)
- {
- if (m_pTarget && m_pfnSelector)
- {
- (m_pTarget->*m_pfnSelector)(m_fElapsed);//看这里
- }
- ...................
- }
- }
- else
- {//advanced usage
- ...............
- if (m_bUseDelay)
- {
- if( m_fElapsed >= m_fDelay )
- {
- if (m_pTarget && m_pfnSelector)
- {
- (m_pTarget->*m_pfnSelector)(m_fElapsed);//看这里
- }
- ...............
- }
- }
- else
- {
- if (m_fElapsed >= m_fInterval)
- {
- if (m_pTarget && m_pfnSelector)
- {
- (m_pTarget->*m_pfnSelector)(m_fElapsed);//看这里
- }
- .......................
- }
- }
- ..............
- }
- }
- }
你可能还不知道(m_pTarget->*m_pfnSelector)(m_fElapsed);这个什么意思,没事,重新看看HelloWorldScene.h调用update函数时的代码:
this->schedule(schedule_selector( HelloWorld::update));
下面是schedule的实现(上面已提过)
- void CCNode::schedule(SEL_SCHEDULE selector)
- {
- this->schedule(selector, 0.0f, kCCRepeatForever, 0.0f);
- }
- void CCNode::schedule(SEL_SCHEDULE selector, float interval, unsigned int repeat, float delay)
- {
- CCAssert( selector, "Argument must be non-nil");
- CCAssert( interval >=0, "Argument must be positive");
- m_pScheduler->scheduleSelector(selector, this, interval , repeat, delay, !m_bRunning);//看这里
- }
最后一步,就是看m_pScheduler->scheduleSelector(selector, this, interval , repeat, delay, !m_bRunning)和
(m_pTarget->*m_pfnSelector)(m_fElapsed)有什么关系,其中selector是函数指针,指向
HelloWorld::update,不懂的可以看
schedule_selector的定义。
下面是scheduleSelector函数的实现
- void CCScheduler::scheduleSelector(SEL_SCHEDULE pfnSelector, CCObject *pTarget, float fInterval, unsigned int repeat, float delay, bool bPaused)
- {
- ................
- tHashTimerEntry *pElement = NULL;
- HASH_FIND_INT(m_pHashForTimers, &pTarget, pElement);
- if (! pElement)
- {
- ...........................
- }
- else
- {
- CCAssert(pElement->paused == bPaused, "");
- }
- if (pElement->timers == NULL)
- {
- pElement->timers = ccArrayNew(10);
- }
- else
- {
- ....................
- }
- CCTimer *pTimer = new CCTimer();
- pTimer->initWithTarget(pTarget, pfnSelector, fInterval, repeat, delay);//看这里
- ccArrayAppendObject(pElement->timers, pTimer);
- pTimer->release();
- }
重点看pTimer->initWithTarget(pTarget, pfnSelector, fInterval, repeat, delay),其中pTarget就是CCNode(对应
HelloWorld类)对象本身,而pfnSelector就是刚才说的函数指针,指向
HelloWorld::update,下面是initWithTarget的函数实现
- bool CCTimer::initWithTarget(CCObject *pTarget, SEL_SCHEDULE pfnSelector, float fSeconds, unsigned int nRepeat, float fDelay)
- {
- m_pTarget = pTarget;
- m_pfnSelector = pfnSelector;
- .....................
- return true;
- }
由此,我们解决刚才提出的问题:
看看m_pScheduler->scheduleSelector(selector, this, interval , repeat, delay, !m_bRunning)和
(m_pTarget->*m_pfnSelector)(m_fElapsed)有什么关系,其中selector是函数指针,指向
HelloWorld::update。
答案就在上面,再次啰嗦说明一下,pTarget就是CCNode(对应
HelloWorld类)对象本身,而pfnSelector就是刚才说的函数指针,指向
HelloWorld::update,也就是说,
HelloWorld类定义的void update(float dt)被调用了,而且是在整个主循环里面
CCDirector::sharedDirector()->mainLoop()。
(5)结束语:
整个cocos2d-x的框架源码就是这样,具体一些细节还未涉及,但我们已经理解程序一开始从main函数,然后怎么样一步步调用到我们自己是HelloWorld.h里面自己定义实现的函数,特别是update函数,同时也了解了cocos2d-x的一些内存管理机制和回调机制,也在这个框架代码里面,还有整个框架是怎么实现帧数控制的。
用《一代宗师》里面的一句话来概括这次框架源码研读之旅吧——”念念不忘,必有回响“。