Steven D'Aprano wrote:
Terry Reedy wrote:
In fact, one reason to subclass a class is to change the initialization api.
That might be a reason that people give, but it's a bad reason from the perspective of interface contracts, duck-typing and the LSP.
Only if you're going to pass the class off to something as a factory function. Note that having a different constructor signature is *not* an LSP violation for *instances* of a class. The constructor is not part of the interface for instances, only for the class itself. In practice, it's very common for a class to have a different constructor signature from its base class, and this rarely causes any problem. IPython is simply making a dodgy assumption. It gets away with it only because it's very rare to encounter subclasses of list or tuple at all. -- Greg -- http://mail.python.org/mailman/listinfo/python-list
