Guido van Rossum wrote:
Sooo should (for 'generator' in objects that claim to be in __builtins__ but aren't), 1) 'generator' be added to __builtins__ 2) 'generator' be added to types.py and its __module__ be set to 'types' 3) 'generator' be added to <newmodule>.py and its __module__ be set to '<newmodule>' (and a name for the module chosen)
I guess (1).
[Note: it should be added to __builtin__, not __builtins__ -- the two are related but __builtin__ is the module and __builtins__ is a secret attribute that references the module or its __dict__ for purposes of sandboxing. [Brett C]
The problem I see with explicitly adding this stuff to __builtins__ is that tab-completion in the interpreter is suddenly going to have all of this extra stuff that people are not going want to really see that often.
There's already lots of stuff there that's rarely used.
The other point I would like to make is that almost everything in __builtins__ is either an exception, a singleton (thinking of True and False), or a function (even thinking of list and dict since you can use their factory methods).
But list and dict are *not* functions, they are types, and they can be used for subclassing and type checking too. These are the precedent for placing generator etc. in __builtin__.
The only exceptions I can think of are __name__ and that is just part of the design. Throwing in generator and any of the other types that are defined by the built-in types will go against all of this unless we create factory methods for them which is not really desired since they are returned only in certain situations.
They should not be made factory functions if they aren't already.
I personally prefer option 2.
Bah. The types module shouldn't be the "home" for any types -- it's just a collection of convenient aliases, and mostly they have the wrong names. It should be deprecated. Also note that at least 6 of these "anonymous" built-in types currently have another alias in the 'new' module, which means that they *are* factories already. It would be nice if that module could finally be deprecated. --Guido van Rossum (home page: http://www.python.org/~guido/)