Sooner or later authors and maintainers of libraries change public interfaces of their creations. Usually one of the two approaches is taken:
1. Outright breaking change 2. Soft deprecation later followed by [1] While [1] is perfectly suitable for libraries with limited audience, [2] is what popular libraries have to do (to remain popular). Python's stdlib contributed to the process via PEP 230 and, recently, PEP 565. The `warn` function and subclasses of Warning are omnipresent. However, when it comes to practical application it's still up to a developer to write utility methods and classes. I propose that Python should extend the warnings module with a set of utilities to cover basic needs as seen across the industry. The extension should include: 1. A decorator for functions 2. A decorator for classes 3. A context manager Function Decorator A function can be deprecated: - Via a rename when an author decides there is a better name - Via a change of its interface - By becoming obsolete Therefore the job of the decorator is to modify the behavior such as upon a call: - A warning is issued - Optionally an aliased (new) function is called - Optionally both positional and keyword arguments are mapped to satisfy new interface; might be followed by an extra warning for each remap Class Decorator A class can be deprecated: - Via a rename when an author decides there is a better name - By becoming obsolete Therefore a job of the decorator is to ensure that: - A warning is issued when class is either instantiated or used as a base class - If aliased, subclasses of a deprecated class (from the end user perspective) pass `issubclass` and `isinstance` checks against an aliased new class - If aliased, subclasses of a new class pass `issubclass` and `isinstance` checks against a deprecated class Context Manager Context manager is useful when behavior of a particular function changes without visible changes to its name or interface. E.g. when author realizes (too late) that arguments are mutually exclusive. Instead of introducing a breaking change outright, better solution is to wrap that particular case with a warning and consistent behavior. Each of these utilities should allow to specify a condition via a callable. E.g. when deprecation only makes sense for certain platforms or python interpreter versions. I think most if not all can be implemented using existing runtime and stdlib. But before diving into code, what do you think of this idea overall?
_______________________________________________ Python-ideas mailing list Python-ideas@python.org https://mail.python.org/mailman/listinfo/python-ideas Code of Conduct: http://python.org/psf/codeofconduct/
