[PyQt] Unsatisfactory behaviour of QSettings

[Baz Walter]

Hello

When using the v1 API with QSettings, it's easy to get invariant output 
- even with empty, or single-element lists:

 >>> s = QSettings('test', 'test')
 >>> s.setValue('key', [])
 >>> s.sync()
 >>> v = s.value('key')
 >>> v.isNull()
True
 >>> list(v.toStringList())
[]
 >>>
 >>> s.setValue('key', ['a'])
 >>> s.sync()
 >>> list(v.toStringList())
[PyQt4.QtCore.QString(u'a')]

But things get really tricky if you try to specify the type and/or use 
the v2 API:

 >>> s.setValue('key', [])
 >>> s.sync()
 >>> v = s.value('key', [], 'QStringList')
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
TypeError: unable to convert a QVariant of type 0 to a QMetaType of type 11
 >>>
 >>> s.setValue('key', ['a'])
 >>> s.sync()
 >>> v = s.value('key', [], 'QStringList')
 >>> list(v.toStringList())
[<PyQt4.QtCore.QStringList object at 0x1a5c500>]

Raising a TypeError here seems a bit trigger-happy. I suppose the reason 
why it's happening is because QVariant.canConvert will always return 
False if the variant is null, even though the conversion is actually 
valid. According to the Qt docs, this is done for "historical reasons" - 
but even so, it's still there in Qt5.

The list-within-list for single-element lists is much harder to explain, 
though - could it be a PyQt bug?

Whatever the causes, these two issues make it very awkward to get 
invariant output when using the v2 API.

Is it too late to try to improve things?

-- 
Regards
Baz Walter