some work on type hint stubs

[Kyle Altendorf]

On 2020-10-09 17:14, Phil Thompson wrote:
> On 09/10/2020 20:04, Kyle Altendorf wrote:
>> Hi Phil,
>> 
>> There's been some work over in PyQt5-stubs [0] recently.  It is now
>> passing mypy --strict (with 180 reviewed ignores) and we are working
>> through the complaints from stubtest (mostly enumeration __new__
>> signatures are left).  Not that automated generic checks catch
>> everything, but they are pieces of the puzzle.  Perhaps this work
>> could be useful as a reference for some changes to the official stub
>> generation.  If you would be interested in any help enhancing the
>> official stubs, let us know how.
>> 
>> Cheers,
>> -kyle
>> 
>> 
>> [0] https://github.com/stlehmann/PyQt5-stubs
> 
> I wasn't aware of this. It's on my list for SIP v6 to review the
> support for stubs as it hasn't much changed much since it was first
> implemented.

Since you mention SIP v6, would you expect to make corrections for 
future PyQt5 releases?  Or just PyQt6?  I'll admit I haven't been 
closely following plans.

> I'm happy to take bug reports (and patches!) to fix any issues - the
> automatically generated stubs should be correct without any need to
> maintain separate versions.

I generally agree and my intent on working on PyQt5-stubs was to try to 
explore the problems and solutions without worrying about learning the 
generation at the same time.  Then there would be a reference target to 
update the generation towards, whether I worked on it or someone else.  
That said, there's enough going on here and enough things that mypy 
doesn't support at present that there's room for at least some opinions 
in certain areas as opposed to simply being correct and incorrect.  But 
anyways...  Some of the stuff seems like it will just have to have a 
mypy plugin.  :|

https://github.com/stlehmann/PyQt5-stubs/pulls?q=is%3Apr+is%3Aclosed

The merged pull requests could constitute a list of 'bug reports'.  I 
tried to keep some semblance of self control and keep them granular.  
Mostly.  I'll also try to take a stab at learning about the existing 
generation mechanisms.  As changes are made upstream we can do catch up 
adjustments in PyQt5-stubs to keep track of the remaining diff.

Side note, I've also got a PR in to qtpy to 'support' type hints (they 
mostly just have to turn it on so the underlying hints from PyQt and 
PySide can pass through, then a little helper for configuring mypy).  Of 
course, PySide also has various hinting issues, many similar ones, some 
that I have reported.

https://github.com/spyder-ide/qtpy/issues/216

Cheers,
-kyle