Hi, people. Since a few days, I'm working at a Python rewrite of a difficult and complex C application, written by a "power" programmer years ago, who left the company before I came in. The rewrite goes extremely well, Python legibility and facilities are such that I could improve the angle of attack for the problem, making it appear much, much easier. The C code mallocs a lot, and the Python rewrite is memory intensive. Since it handles hundreds of thousands of objects, I thought it was a good case for using `__slots__' and indeed, the memory savings are interesting. Deriving from built-in types for a good amount of objects (for a limited amount of types), whenever appropriate, is also welcome. In one module, sub-classing is frequent, and if I understand well how it works, a `__dict__' is used in sub-types unless `__slots__ = ()' is explicitely included in the sub-type, and I decided to spare that `__dict__' whenever possible in that particular module. So, just after the global `__metaclass = type' at the beginning of the module, I added a global `__slots__ = ()', with the clear intent of overriding that `__slots__' assignment only for those particular types needing slots. In this module of the application, I do not mind having to restrain all types so none uses a `__dict__'. But it does not seem to work that way. Could it be pondered that, the same as `__metaclass__' may have a global value, `__slots__' may also have? One objection I see is that, while __metaclass__ may be redefined to the classic class metatype for a specific type, a global `__slots__' could not be arbitrarily defeated in a specific type. Nevertheless, in the precise case I have, it would have been a bit useful. I do not mind much having to repeat `__slots__' in each and every class and sub-class for this module, yet I thought about sharing the above thoughts with Python developers, in case they would like to level the said difference. It would be less surprising if `__metaclass__' and `__slots__' could equally be implied globally. Or maybe? :-) -- François Pinard http://www.iro.umontreal.ca/~pinard
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4