The signal/slot mechanism in Qt, is a static mechanism. The classes have to be preprocessed by the moc compiler.
Now I want to create signals and slots dynamically at run-time.
I already have a working solution, but it feels to me like a hack, although I am using publicly available methods.
This is the code for dynamic slots:
bool DynamicQObject::connectDynamicSlot(const QString &objectName, QObject *pSourceObject, QMetaMethod signalMethod)
{
QByteArray slotName = signalMethod.name().prepend("on").append("(");
QStringList parameters;
for (int i = 0, j = signalMethod.parameterCount(); i < j; ++i)
{
parameters << QMetaType::typeName(signalMethod.parameterType(i));
}
slotName.append(parameters.join(",")).append(")");
QByteArray theSignal = QMetaObject::normalizedSignature(signalMethod.methodSignature().constData());
QByteArray theSlot = QMetaObject::normalizedSignature(slotName);
if (!QMetaObject::checkConnectArgs(theSignal, theSlot))
{
return false;
}
int signalId = pSourceObject->metaObject()->indexOfSignal(theSignal);
if (signalId < 0)
{
return false;
}
int slotId = slotIndices.value(theSlot, -1);
if (slotId < 0)
{
slotId = slotList.size();
slotIndices[theSlot] = slotId;
slotList.append(createSlot(theSlot, objectName, signalMethod));
}
return QMetaObject::connect(pSourceObject, signalId, this, slotId + metaObject()->methodCount());
}
As you can see, I make intensive use of the QMetaObject, and particularly the index of the slots (method count).
The code for dynamic signals is comparable.
Now my question is: how future proof is this solution, especially because I assume that the index must be at least one bigger than the methodCount()?
Now my question is: how future proof is this solution, especially because I assume that the index must be at least one bigger than the methodCount()?
It should work as of now.
As for future proof... may be. The snippet is using unsupported features which means those might break at any point in the future. It is likely though it will continue working.