HiDPI monitors

[ullix]

@Elvis: duly noted! But allow me to postpone for a few days to allow 
that little success ;-)


Am 16.01.21 um 14:28 schrieb Elvis Stansvik:
> Den lör 16 jan. 2021 kl 14:26 skrev Elvis Stansvik <elvstone at gmail.com>:
>> Nice! Just a small comment about
>>
>>> MySetting = 100 / int(QScreen.devicePixelRatio(app.primaryScreen()))
>> Here you are assuming that the application is on the primary screen.
>> In a multi-monitor setup, it may be on another screen with a different
>> device pixel ratio. It's possibly a better idea to call
>> QPaintDevice::devicePixelRatio() on the widget that will hold your
>> matplotlib plots.
>>
>> Also, the application may be moved between screens, which you should
>> ideally listen for (there are some Qt mechanisms for this I believe,
>> screenChanged signal, though I've found it a bit unreliable at times),
>> and if it happens, re-create your matplotlib plots with the correct
>> DPI passed in (calculated from the new screen the widget is on), and
>> also re-create any pixmaps you may have created in terms of physical
>> pixels (if you have any like that).
> I see now that you said "and the user says he can even move it back
> and forth between an 8K and a FullHD monitor without distortion except
> for a wobble in the transition !!?", so perhaps matplotlib takes care
> of this on its own once the plot has been shown, which is great.
>
> Elvis
>
>> How much of this you want to bother with is up to you of course :) You
>> may decide it's enough if the user has to restart the application if
>> he/she wants it on another screen.
>>
>> HiDPI really is a pain :)
>>
>> Cheers,
>> Elvis
>>
>> Den fre 15 jan. 2021 kl 17:34 skrev ullix <ullix at urkam.de>:
>>> My struggle with HiDPI monitors has come to an at least workaround solution thanks to your responses, so many thanks!
>>> I'll try to summarize:
>>>
>>> PyQt(5) applications will behave well with this command:
>>>
>>> QtWidgets.QApplication.setAttribute(QtCore.Qt.AA_EnableHighDpiScaling, True)
>>>
>>> Whether this next command will also be needed depends on your explicit (perhaps implicit) use of pixmaps with HiDPIs available
>>> I had no benefit from it
>>>
>>> QtWidgets.QApplication.setAttribute(QtCore.Qt.AA_UseHighDpiPixmaps, True)
>>>
>>> Unfortunately, matplotlib is separate from this and needs its own adaptation of settings. This can be done by setting the 'dpi=' parameter in every call to figure:
>>> matplotlib.pyplot.figure(num=None, figsize=None, dpi=MySetting, .... )
>>>
>>> But MySetting is a bit tricky: it basically must reverse the scaling. The default dpi is: MySetting = 100. So, when your scaling is 200%, then set MySetting=50. Is your scaling 300%, then set MySetting=33.
>>>
>>> I don't know if there is a single command which could set dpi for the session?
>>>
>>> Thankfully the "devicePixelRatio" of QScreen basically has Scaling/100, i.e. the values 1, 2, 3 for the above examples. Therefore you can set:
>>> MySetting = 100 / int(QScreen.devicePixelRatio(app.primaryScreen()))
>>>
>>> In the rare case that you do not have several 4k and 8k monitors to test with :-( you can use Xephyr as suggested by VA.
>>> Though when I tried the examples refernenced by him  https://indigo.re/posts/2020-05-18-hidpi-testing.html each one crashed on my system! But in my need I came back and just tried the small script and it worked! Script: https://gitlab.com/hydrargyrum/attic/blob/master/xephyr-run-cmd/xephyr-run-cmd
>>>
>>> I can now start my program and run it in proper scale with this command:
>>> ./xephyr-run-cmd -dpi 200 -screen 4000x2000 -- ./geigerlog
>>>
>>> And lastly a screenshot on a real 8k monitor on Win10:
>>> Full WidthxHeight is 2732x1443 (open it);
>>> and the user says he can even move it back and forth between an 8K and a FullHD monitor without distortion except for a wobble in the transition !!?
>>>
>>> Thanks again,
>>>
>>> ullix
>>>
>>>