From pykde  Fri Dec 22 10:57:41 2023
From: Phil Thompson <phil () riverbankcomputing ! com>
Date: Fri, 22 Dec 2023 10:57:41 +0000
To: pykde
Subject: Re: Issues when promoting QPdfView in Designer
Message-Id: <42cb7b8d499ee51adec9f6007c8b45a7 () riverbankcomputing ! com>
X-MARC-Message: https://marc.info/?l=pykde&m=170324268315084

On 21/12/2023 18:57, Maurizio Berti wrote:
> Some time ago I realized that the QPdfView and QPdfDocument classes
> [re]introduced since Qt6.4 have the parent argument as mandatory, even 
> if a
> real parent is actually required (using None is perfectly valid).
> I've never used these objects, except for knowledge purposes, so I 
> didn't
> mind too much.
> 
> Yesterday I came across this SO post:
> https://stackoverflow.com/q/77692836/2001654
> 
> In this case the problem becomes more important: using QPdfView as a
> promoted widget in Designer results in a TypeError because the parent
> argument is keyworded, but even if the positional named argument is 
> still
> "parent", it's not recognized as a standard python function would.
> 
> I know that the issue mostly comes from the "wrong signature" in Qt and
> PyQt just does what its implementation is expected to do, and I also 
> filed
> a report (QTBUG-120313 <https://bugreports.qt.io/browse/QTBUG-120313>).
> 
> In the meantime, I proposed a couple of possible workarounds: patching
> instantiate() in qobjectcreator.py, or using a custom QPdfView subclass 
> as
> a promoted widget instead of the default one.
> 
> Still, I'd like to know the exact reason for which the parent argument
> isn't accepted if keyworded, since the apparent signature should allow
> that. I know that PyQt classes do some "magic" under the hood whenever 
> a
> new instance is created, but I'm curious about these aspects and the
> difference with the normal, expected behavior.
> 
> I'd also suggest to provide an official fix for this issue, considering
> that there's no guarantee that the Qt class constructors will be fixed
> anytime soon (and there would still be problems with previous Qt 
> versions).
> Nonetheless, I realize that "officially" fixing this may be difficult 
> or
> inconsistent and would potentially break some aspects (even 
> semantically),
> in that case a note in the "Using Designer" documentation should 
> probably
> be added.
> 
> Thank you,
> MaurizioB

To answer the question about keyword arguments, PyQt uses the "Optional" 
mode as described in...

https://www.riverbankcomputing.com/static/Docs/sip/annotations.html#function-annotation-KeywordArgs

...so that keywords cannot be used for any argument that is not 
optional.

Regarding a proper fix, special casing QPdfView in the code isn't 
acceptable but there are a couple of things that can be exploited. First 
is that support for a non-optional parent is already there - see 
_setupObject() in uiparser.py. Second is the widget plugin mechanism 
(see the widget-plugins directory). It should be possible to extend the 
latter to configure the former.

Phil