Re: [PATCH 04/37] qapi: move generator entrypoint into module

[Markus Armbruster <armbru@redhat.com>]

John Snow <jsnow@redhat.com> writes:

> On 9/16/20 7:54 AM, Markus Armbruster wrote:
>> John Snow <jsnow@redhat.com> writes:
>> 
>>> As part of delinting and adding type hints to the QAPI generator, it's
>>> helpful for the entrypoint to be part of the module, only leaving a very
>>> tiny entrypoint shim outside of the module.
>>>
>>> As a result, all of the include statements are reworked to be module-aware,
>>> as explicit relative imports.
>> Should this be split into two commits, one for each of the
>> paragraphs
>> above?
>> 
>
> Hmm ... I hadn't considered it was possible, but actually ... I guess
> I can split those out, yeah.
>
>> PEP 8 recommends absolute imports, with one exception:
>>      However, explicit relative imports are an acceptable
>> alternative to
>>      absolute imports, especially when dealing with complex package
>>      layouts where using absolute imports would be unnecessarily verbose:
>>          from . import sibling
>>          from .sibling import example
>>      Standard library code should avoid complex package layouts and
>>      always use absolute imports.
>> Do you think it covers your use of relative imports?
>> 
>
> Admittedly I am going by trial and error; my experience is that the
> relative imports behave the nicest.
>
> There is a historical fear of explicit relative imports because they
> are "new" and years of Python2 compatibility

Having to struggle with that for years was bound to damage brains.

>                                              rendered many afraid of
> them. It is advice safely ignored in my opinion.
>
> Using absolute imports (e.g. from qapi.sibling import foo) gets messy
> in virtual environments when you have *installed* the module in
> question: it becomes ambiguous as to *which* qapi you meant: the one
> in this folder, or the one installed to the environment?
>
> Python, mypy, pylint, flake8 etc disagree sometimes, or can get
> confused; thinking there are two copies of each module.
>
> Just my experience that relative imports seem to give me the least trouble.
>
>>> This is done primarily for the benefit of python tooling (pylint, mypy,
>>> flake8, et al) which otherwise has trouble consistently resolving
>>> "qapi.x" to mean "a sibling file in this folder."
>> Can you give me an example of some tool having troube?
>> 
>
> I'd have to code up some examples. I have some hobby code unrelated to
> QEMU where I struggled to get flake8, mypy, and pylint all cooperating 
> with an import regime until I gave up and used explicit relative imports.

I'd make the switch when we actually do run into trouble.

But I'm willing to take your advice and switch now.

[...]