> On 15 Apr 2023, at 12:07, Volker Hilsheimer <volker.hilshei...@qt.io> wrote:
>
>> On 14 Apr 2023, at 23:11, Thomas Sevaldrud <tho...@silentwings.no> wrote:
>>
>> Hi, in my code I have a hierarchical QMultiMap of QVariants, where a map
>> value can be a new multimap of variants. This does not compile anymore in Qt
>> 6. I.e. something like this:
>>
>> QMultiMap<QString, QVariant> varMap;
>> QMultiMap<QString, QVariant> subMap;
>> varMap.insert("key", subMap); // Compile error here, since there is no
>> conversion to QVariant
>>
>> This worked in 5.15, but not in 6.5
>>
>> If I change it to QMap it works, but I need multimap for this. Is this
>> change by design, or is it an omission? Any ideas for a workaround?
>>
>> Cheers,
>> Thomas
>
>
> In Qt 5, QMultiMap was a subclass of QMap, so implicitly constructing a
> QVariant from a QMultiMap was possible because it was possible from a QMap.
> That happened to work because QMultiMap just added API on top of a QMap, and
> QMap’s data structure could handle multiple values for the same key. So no
> slicing occurred. However, if you run:
>
> #include <QtCore>
>
> int main(int argc, char *argv[])
> {
> QMultiMap<QString, QVariant> varMap;
> QMultiMap<QString, QVariant> subMap;
> varMap.insert("key", subMap);
> qDebug() << varMap.value("key").typeName();
>
> return 0;
> }
>
> then you would get QVariantMap (ie a QMap<QString, QVariant>), not QMultiMap.
> So modifying the value you got out of the map would not give you multi-map
> semantics.
>
>
> In Qt 6, QMultiMap is not a QMap subclass anymore (and that was a design
> decision), and since it’s such a rarely used type there is also no implicit
> constructor or conversion operator, and no built-in metatype defined for
> QMultiMap.
>
> But since QVariant can be constructed from custom types via the fromValue
> template, you can use:
>
> varMap.insert("key", QVariant::fromValue(subMap));
>
> and now you can’t get it out as a QVariantMap anymore, and the code above
> will print “QMultMap<QString,QVariant>”, like it should.
>
> Volker
>
Cool, thanks! That makes sense. We probably never had any submaps in this
structure with non-unique keys, so we never got this problem. This is of course
much better.
Thanks again!
- Thomas
_______________________________________________
Interest mailing list
Interest@qt-project.org
https://lists.qt-project.org/listinfo/interest
