Author zzzeek
Recipients brett.cannon, hroncok, lukasz.langa, ncoghlan, pablogsal, serhiy.storchaka, steve.dower, vstinner, zzzeek
Date 2019-05-16.19:10:58
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1558033859.0.0.417142010317.issue36751@roundup.psfhosted.org>
In-reply-to
Content
> A deprecating warning doesn't hurt: you are still able to run your code.

this is very much untrue in modern testing environments, as it is common that test suites fail on deprecation warnings, especially in libraries, to ensure downstream compatibility.  My observation is that as Python 3 has been generally a lot more aggressive about new language features than things used to be in the py2k days, a lot more of these deprecations have been coming through.  

I've looked at pep 570 and this new syntax does not affect the four libraries for which I have to rewrite getfullargspec, so while I'm looking a bit at things like the "Signature" backport for py2.7 (funcsigs), the pure-Python complexity of Signature is very hard to take on as well as a py2k-only dependency so I'm still leaning towards producing the most minimal getfullargspec() I can use that continues to work for "traditional" Python signatures.

i understand that Signature is "future-proof", but it is heavy on the pure Python and object creation and does not provide for any simple py2k/py3k cross-compatibility solution.   I'd very much prefer that Signature at least be written in C for a few major versions, as well as the community has had time to move further python-3-only,  before getfullargspec is so aggressively deprecated.
History
Date User Action Args
2019-05-16 19:10:59zzzeeksetrecipients: + zzzeek, brett.cannon, ncoghlan, vstinner, lukasz.langa, serhiy.storchaka, steve.dower, hroncok, pablogsal
2019-05-16 19:10:59zzzeeksetmessageid: <1558033859.0.0.417142010317.issue36751@roundup.psfhosted.org>
2019-05-16 19:10:58zzzeeklinkissue36751 messages
2019-05-16 19:10:58zzzeekcreate