As a rule, no incompatible API changes are allowed during a minor release of numpy. By incompatible API change, we mean:
- change of function/Class/etc... name
- change of order of arguments
- change of default value for an argument
Adding arguments is okay, though.
Deprecated uses should issue a DeprecationWarning using the standard library's warnings module. There are two decorators deprecate() and deprecate_with_doc() in numpy.lib.utils for conveniently deprecating an entire function.
Deprecation warnings should exist for one full minor release cycle before the deprecated features are removed, certainly no less and preferably no more. For example, if a function is deprecated in numpy 1.1.0, it will remain so in all 1.1.x releases, but it can and should be removed in 1.2.0.