Re: [Pyobjc-dev] Namespaces, intended behavior?

[Ronald O. <ron...@ma...>]

That is correct. There is another reason for that: the lazy loader I mentioned.  Apple’s frameworks expose a lot of names and * imports force PyObjC to create Python versions of all of them which uses memory and more importantly takes significant time. 

Ronald

--
On the road, hence brief. 

Op 5 apr. 2018 om 18:43 heeft Just van Rossum <jus...@gm...> het volgende geschreven:

> 
>> On 05 Apr 2018, at 18:12, Ronald Oussoren <ron...@ma...> wrote:
>> 
>> 
>> 
>>> On 5 Apr 2018, at 17:39, Just van Rossum <jus...@gm...> wrote:
>>> 
>>> Hi,
>>> 
>>> I am wondering, is the following intended behavior?
>>> 
>>> -------------
>>> import Foundation
>>> 
>>> class TestMyCustomObject(Foundation.NSObject):
>>>  pass
>>> 
>>> import AppKit
>>> # it's suddenly available via AppKit:
>>> print(AppKit.TestMyCustomObject)
>>> # even via import *:
>>> from AppKit import *
>>> print(TestMyCustomObject)
>>> -------------
>>> 
>>> Thanks,
>> 
>> That is expected behavior and is a side effect of lazy loading of frameworks.
>> 
>> In the past I have attempted to expose only the classes that belong in a framework (by exposing only those classes located in the correct framework bundle), but that code turned out to be too slow (because the Cocoa method I used for that is slow), so that got dropped a long time ago. When I started on the lazy loader I didn’t even try to make it guess whether or not a class should be loadable through the current framework wrapper.  This is not ideal, but good enough.
> 
> Thanks for the clarification. That does mean that one should pretty much never ever do
> 
>  from SomeFramework import *
> 
> Despite the fact that most frameworks have decent prefixes for their sybols.
> 
> Just
>