Robinhilliard Pipes Save

Elixir–Style Pipes for Python

In the Elixir programming language the |> pipe operator allows you to chain together multiple function calls so that this:

c(b(a(1, 2), 3, 4))

can be written more readably as:

1 |> a(2) |> b(3, 4) |> c()

All the pipe operator does is pass its left operand as the first argument of the right operand, so that a |> b(...) becomes b(a, ...).

Various pipe implementations in Python to date allow a list of functions to be applied to an initial value, but do not support the partial, missing first argument syntax of Elixir.

This library provides a function decorator that causes Python >> right shift operators within the function to act exactly like Elixir pipes:

from pipeop import pipes

def add(a, b):
    return a + b
    
def times(a, b):
    return a * b
    
@pipes
def calc():
    print 1 >> add(2) >> times(3)  # prints 9

Functions can have any number of arguments:

def add3(a, b, c):
    return a + b + c
    
@pipes
def calc():
    print 1 >> add3(2, 3)  # prints 6

In Elixir libraries the first argument of a function is chosen with pipes in mind but this is (obviously) not the case in Python - for instance the enumerable args of map and reduce are first in their Elixir equivalents but last in Python. For this reason I've also redefined the left shift operator << to append it's left operand to the list of call arguments of the right operand:

@pipes
def my_pow():
  print 2 >> pow(3)  # prints 8
  print 2 << pow(3)  # prints 9

You can drop the braces for functions or lambdas (enclosed in braces) with a single argument:

@pipes
def sum(self):
    print [1, 2, 3] >> sum  # prints 6
    print 1 >> (lambda x: x + 1)  # prints 2

In Elixir pipes are often laid out one per line. In Python you need brackets to do the same thing without line continuations, but it still looks pretty neat:

@pipes
def pretty_pipe():
    print (
        range(-5, 0)
        << map(lambda x: x + 1)
        << map(abs)
        << map(str)
        >> tuple
    )  # prints ('4', '3', '2', '1', '0')

The decorator can also be applied to a class to decorate all the methods in that class.

Normally there should be a small amount of processing overhead on the first time the function is defined due to the function being recompiled and cached. Otherwise there should be no difference to the performance of the conventionally nested call code.

This is initial alpha code. It has been tested on Python 2.7.14 and 3.6.5 using simple functions. Source line attributes are preserved so debuggers should be able to follow the code as it executes. Pull requests and bug reports gratefully accepted.

Robin Hilliard

PS: Thanks to https://github.com/Stvad for submitting the first issue with some great suggestions.