FuPy: Functional Programming in Python, for Education

Overview

• Documentation
• For some examples, see demo.py.

Python offers (limited) support for functional programming:

• functions are first-class citizens;
• functions can take functions as arguments and can return functions
• lambda expressions for anonymous functions

FuPy aims to overcome some limitations, by adding:

• basic types and type constructors
  • Empty type without values
  • Unit type with value unit
  • Either[A, B] (disjoint sum) type
  • Both[A, B] (product) type (tuple)
  • Func[A, B] function space type
• function composition and other function combinators
  • @ for head-to-tail composition, also written as ⚬
  • & for split, also written as △
  • | for case, also written as ▽
  • * for product (functor over product/tuple type), also written as ⨯
  • + for sum (functor over sum/Either type), also written as +
  • ** for iterated composition, also written with a superscript
  • ^ for functorial exponentiation
• auto (un)packing of arguments in case of functions with no/multiple arguments
• predefined common functions:
  • id_, const, left, right, guard, first, second
  • curry, uncurry, flip, ap,
  • compose, split, case_, fplus, ftimes, fpower, fpower_left
• operator sections
• lazy expressions (suspended computations)
• printable function and lambda expressions
• evaluation tracing
• inductive and co-inductive types
  • Functor, fmap
  • Fix, fix
  • cata (catamorphisms, folds)
  • ana (anamorphims, unfolds)

Main classes:

• Func, for composable, printable, and traceable functions with auto-(un)packing of arguments.
• OperatorSection, for operator sections
• Lazy, for lazy expressions.

Not intended for industrial use:

• there are performance penalties in terms of memory and execution overhead

Notes:

• Type hints do not all verify (but it works). The Python type system is too limited (we need Higher-Kinded Types, HKTs).
• Binding strength of function combinators are as applied by Python, doesn't correspond to the theory.

Future work

• Applicatives and monads

Installation

pip install FuPy

Development (for developers)

• Build:

  python -m build
  

• Test:

  pytest
  

• Build documentation: In docs/

  make clean
  make html
  

• Upload to PyPI:

  twine upload dist/*
  

Authors and acknowledgment

• Tom Verhoeff (Eindhoven University of Technology, Netherlands)

License

MIT License

Project status

• Under development, but already usable
• Documentation:
  • incomplete
  • Sphinx version still has issues
• Test cases: incomplete
• Functionality:
  • functions only print in mixed Math/Python notation
  • level of detail in tracing cannot be selected
  • limited form of laziness
• Issues