NSObject PerformSelector问题

Question

好吧，还有一个“我只是不知道发生了什么”的问题：

当我第一次调用下面的getFullWL()时，我得到了所有预期的值。但是，随后的每个调用都返回nan而不是真值(-nan(0x400000000)在XCode或其他方面)。

此外，如果我将“调试”行放在SFLog中，则值打印为nan，但返回正确的值！如果我注释掉了SFLog (这只是NSLog的#定义，取决于我设置的级别--没有什么特殊)，那么这个值不会被isnan()在View中捕获，而是在Window中被选中为isnan()

Window和View坐在一个线程上，而DataModule和DCMPix坐在另一个线程上(我相信主线程)。DCMPix还从核心数据对象(我相信是fImage)获得一些数据。

void Window::PopUp()
{
    // Grab the View from the Pipe Thread and cast it to my local namespace subclass View
    View *view = static_cast< View* >(getPipe()->getView(event.context.view));

    float fullww = view->getFullWW();
    float fullwl = view->getFullWL();
    //SFLog(@"WindowLevel Window %f", wl);

    // set SLider Values
    if (!isnan(fullwl)){
        [[vc winLevel] setMinValue:fullwl];
        [[vc winLevel] setMaxValue:(-1.0f*fullwl)];
    }
}


float View::getFullWL()
{
    float wl = _callPixMethod(@selector(fullwl)).floatValue;
    if ( isnan(wl) ) wl = _wl * 2.0f; //<-- is never detected as NaN here
    //SFLog(@"WindowLevel View %f", wl);  //<-- but is *printed* as NaN here
    return wl;
}

// A Union to "cast" the return from the [object performSelector:] method.
//   Since it returns `id` and (float)(id) doesn't work.
union performReturn {
    id OBJC_ID;
    int intValue;
    float floatValue;
    bool boolValue;
};


performReturn View::_callPixMethod(SEL method)
{
    DataModule* data;
    DataVisitor getdata(&data);
    getConfig()->accept(getdata);
    performReturn retVal;
    retVal.OBJC_ID = data->callPixMethod(_datasetIndex, _datasetID, method);
    return retVal;
}


id DataModule::callPixMethod(int index, std::string predicate, SEL method)
{
    DCMPix *pix =[[getSeriesData(predicate) pixList_] objectAtIndex:index];

    pthread_mutex_lock(&_mutex);

    id retVal = [pix performSelector:method];

    pthread_mutex_unlock(&_mutex);

    return retVal;
}

// DCMPix.m (from API, can't change)
- (float) fullwl
{
 if( fullww == 0 && fullwl == 0) [self computePixMinPixMax];
 return fullwl;
}

- (void)computePixMinPixMax
{
    float pixmin, pixmax;

    if( fImage == nil || width * height <= 0) return;

    [checking lock];

    @try 
    {
        if( isRGB)
        {
            pixmax = 255;
            pixmin = 0;
        }
        else
        {
            float fmin, fmax;

            vDSP_minv ( fImage,  1, &fmin, width * height);
            vDSP_maxv ( fImage , 1, &fmax, width * height);

            pixmax = fmax;
            pixmin = fmin;

            if( pixmin == pixmax)
            {
                pixmax = pixmin + 20;
            }
        }

        fullwl = pixmin + (pixmax - pixmin)/2;
        fullww = (pixmax - pixmin);
    }
    @catch (NSException * e) 
    {
        NSLog( @"***** exception in %s: %@", __PRETTY_FUNCTION__, e);
    }

    [checking unlock];
}

帮助!为什么我不能在Window中得到正确的值？

在调用view->getFullWW()时，即使它遵循相同的执行路径，也不会得到view->getFullWW()。(但是调用@selector(fullww))

不会引发错误或异常。就是那种奇怪的行为。

谢谢!

经过进一步测试后，以下更改将产生所有差异：

float View::getFullWL()
{
    float wl = _callPixMethod(@selector(fullwl)).floatValue;
    if ( isnan(wl) ) wl = _wl * 2.0f; //<-- is never detected as NaN here
    NSLog(@"WindowLevel View %f", wl);  //<-- is *printed* as NaN here
    return wl; //<-- returns expected value
}

float View::getFullWL()
{
    float wl = _callPixMethod(@selector(fullwl)).floatValue;
    if ( isnan(wl) ) wl = _wl * 2.0f; //<-- is never detected as NaN here
    return wl; //<-- returns `nan`
}

经过更多的测试后，如果NSLog()语句发生在分配float fullwl =之前，那么我把它放在哪里并不重要。

void Window::PopUp()
{
    // Grab the View from the Pipe Thread and cast it to my local namespace subclass View
    View *view = static_cast< View* >(getPipe()->getView(event.context.view));

    float fullww = view->getFullWW();
    NSLog("Putting this here, makes fullwl work.  Removing it, stores fullwl as nan");
    float fullwl = view->getFullWL();
    //SFLog(@"WindowLevel Window %f", wl);

Answer 1

孩子，我又在回答我自己的问题了，但答案是.RTFM。

aSelector参数应该标识不带参数的方法。对于返回对象以外的任何内容的方法，请使用NSInvocation。

所以答案是，使用NSInvocation而不是[obj performSelector:]