A relatively sane approach to multiple dispatch in Python.
This implementation of multiple dispatch is efficient, mostly complete, performs static analysis to avoid conflicts, and provides optional namespace support. It looks good too.
See the documentation at https://multiple-dispatch.readthedocs.io/
>>> from multipledispatch import dispatch
>>> @dispatch(int, int)
... def add(x, y):
... return x + y
>>> @dispatch(object, object)
... def add(x, y):
... return "%s + %s" % (x, y)
>>> add(1, 2)
3
>>> add(1, 'hello')
'1 + hello'- Dispatches on all non-keyword arguments
- Supports inheritance
- Supports instance methods
- Supports union types, e.g.
(int, float) - Supports builtin abstract classes, e.g.
Iterator, Number, ... - Caches for fast repeated lookup
- Identifies possible ambiguities at function definition time
- Provides hints to resolve ambiguities when they occur
- Supports namespaces with optional keyword arguments
- Supports variadic dispatch
- Diagonal dispatch
a = arbitrary_type()
@dispatch(a, a)
def are_same_type(x, y):
return True- Efficient update: The addition of a new signature requires a full resolve of the whole function. This becomes troublesome after you get to a few hundred type signatures.
multipledispatch is on the Python Package Index (PyPI):
pip install multipledispatch
It is Pure-Python and depends only on the standard library. It is a light weight dependency.
New BSD. See License file.
- Five-minute Multimethods in Python by Guido
- multimethods package on PyPI
- singledispatch in Python 3.4's functools
- Clojure Protocols
- Julia methods docs
- Karpinksi notebook: *The Design Impact of Multiple Dispatch*
- Wikipedia article
- PEP 3124 - *Overloading, Generic Functions, Interfaces, and Adaptation*