Casting a custom class to QVariant (2020 but still relevant)

Doug Beck

@Christian-Ehrlicher So, QVariantMap only works with data types that aren't based upon QObject? I could really use the ability to map a QString onto a custom class based upon QObject. Ultimately, what I'd like to do is have a QML combo box with a key property and use that key property to look up the model and current index. Is there a better way to do this? I'm still learning the basics of Qt.

mrjj

Hi
To used classes based on QObject, you must store them as pointers.
QObject itself cannot be copied and hence cannot be stored in any container
as plain type. Only via pointers.

Doug Beck

I'm making some progress. I used QMap like this QMap<QString, MyClass*> as a QProperty in my class. Then, in the ComboBox, I used this. I have a property in the QML file called "controlName".

            ComboBox {
                model: MainViewMgr.controls[controlName].nodeList
                width: 132

                onActivated: {
                    MainViewMgr.controls[controlName].currentIndex = currentIndex;
                }

                onModelChanged: {
                    currentIndex = Qt.binding( function(){return MainViewMgr.controls[controlName].currentIndex} )
                }
            }

However, I get this error. Does this mean the Javascript doesn't support using maps?

TypeError: Cannot read property 'nodeList' of undefined

Doug Beck

I have seen some posts on-line that suggest I should be using QVariantMap instead of QMap but I wasn't able to get that to compile with a class based upon QObject.

mrjj

Hi
Yes https://doc.qt.io/qt-5/qtqml-cppintegration-data.html mentions
QVariantMap so that should work.

What error did you get with QVariantMap ?
Also you must also use pointers to "cMyClass"
And it should be supported
https://doc.qt.io/qt-5/qmetatype.html#Type-enum
as it has QMetaType::QObjectStar

Doug Beck

@mrjj The problem is getting myClass into the QVariant. Using the following required me to use Q_DECLARE_METATYPE.

QVariant v = QVariant::fromValue<cMyClass>(MyObject);

However, as soon as I use Q_DECLARE_METATYPE, I get errors about using a deleted function. Without this reference, it compiles just fine. But I can't see a method to insert myClass into QVariantMap without using QVariant::fromValue.

Christian Ehrlicher

Again: you can't copy a QObject! use a pointer...

Doug Beck

@Christian-Ehrlicher There's something I'm not understanding here. Here's my class. When I uncomment the part about Q_DECLARE_METATYPE, I immediately get the compiler error with the use of deleted function.

#include <QObject>
// #include <QMetaType>
#include "PropertyHelper.h"

class cMyClass: public QObject
{
Q_OBJECT
AUTO_PROPERTY(QString, nodeValue)
AUTO_PROPERTY(QString, nodeUnits)
AUTO_PROPERTY(QStringList, nodeList)
AUTO_PROPERTY(qint16, currentIndex)
public:
explicit cMyClass(QObject* parent = nullptr);
explicit cMyClass(QObject* parent, QString controlName, QString value, std::vector<QString> nodeStringVector);
QString controlName() const {return mControlName;}

private:
QString mControlName;
};

// Q_DECLARE_METATYPE(cMyClass);

In my ViewMgr class, I used this as my Q_PROPERTY.

Q_PROPERTY(QVariantMap *myObject READ myObject WRITE setMyObject NOTIFY myObjectChanged)

I guess I'm not sure where I should be using a pointer that I'm not.

Christian Ehrlicher

@Doug-Beck said in Casting a custom class to QVariant (2020 but still relevant):

QVariant v = QVariant::fromValue<cMyClass>(MyObject);

here

Doug Beck

@Christian-Ehrlicher Still get the same errors even without that line of code.

Start of compiler errors*
error: C2280: 'cMyClass::cMyClass:(const cMyClass: &)': attempting to reference a deleted function
compiler has generated 'cMyClass::cMyClass' here
while compiling class template member function 'void *QtMetaTypePrivate::QMetaTypeFunctionHelper<T,true>::Construct(void *,const void *)'
with
[
T=cMyClass
]

see reference to function template instantiation 'void *QtMetaTypePrivate::QMetaTypeFunctionHelper<T,true>::Construct(void *,const void *)' being compiled
with
[
T=cMyClass
]

see reference to class template instantiation 'QtMetaTypePrivate::QMetaTypeFunctionHelper<T,true>' being compiled
with
[
T=cFakeControl
]
End of compiler errors*

I sincerely appreciate your help on this! I have spent much of the last decade using C# so it would not be surprising if I didn't use a pointer correctly. But I'm not sure where the problem is.

Christian Ehrlicher

I don't know what else I can tell. You can't put a QObject into a QVariant, only a pointer to a QObject. Therefore pass a pointer ... c++ basics.

Doug Beck

@Christian-Ehrlicher I get the part about pointers. However, when I copy cMyClass to an empty QML project and don't reference it at all and even remove all of the Q_PROPERTIES from cMyClass I still get the same error as before.

This new project does not use QVariant at all. Nor does it ever refer to cMyClass, which is just an empty class. But, the line Q_DECLARE_METATYPE does cause compiler errors referring to a deleted function. (Which admittedly is just like I'd have if I was trying to use QObject instead of QObject*). So, this has nothing to do QVariant or even my use of cMyClass since I'm not using it at all.

I hope you are saying that Q_DECLARE_METATYPE is completely unnecessary for the use of QVariant.

#include <QObject>
#include <QMetaType>

class cMyClass: public QObject
{
Q_OBJECT
public:
explicit cMyClass(QObject* parent = nullptr);
};

Q_DECLARE_METATYPE(cMyClass);

** Still happens
'cMyClass::cMyClass(const cMyClass &)': function was implicitly deleted because a base class invokes a deleted or inaccessible function 'QObject::QObject(const QObject &)'

mrjj

@Doug-Beck said in Casting a custom class to QVariant (2020 but still relevant):

Q_DECLARE_METATYPE

Hi
https://doc.qt.io/qt-5/qmetatype.html#details
"Any class or struct that has a public default constructor, a public copy constructor, and a public destructor can be registered."
Which means you cant use
Q_DECLARE_METATYPE(cMyClass);

Also I think QObject * is already registered.

Doug Beck

@mrjj Thanks! Clearly that was a dead end. I think I understand what Christian-Ehrlicher was telling me.