Currying iterators in Python

Question

In the following snippet I try to return a function which is returning a generator (weren't they once called iterators BTW?) with the step argument curried.

import math
import numpy

def aequi_ang_step(step):
    def local_it(start, d_alpha, step):
        for alpha in numpy.arange(start, start+d_alpha,(d_alpha < 0) and -step or step):
            if (alpha < 2*math.pi): yield alpha
            else: yield alpha-2*math.pi
    return lambda start, d_alpha: local_it(start, d_alpha, step)

The thing works but as I am returning from a long abstinence back to Python my question is if this is an ok and pythonic way to curry iterators in Python.

[Difference between Python's Generators and Iterators](https://stackoverflow.com/q/2776829/4518341) — wjandrea, Jul 09 '19 at 16:32
I think using `functools.partial` is the Pythonic way to curry an argument, but I don't have much experience with currying myself. — wjandrea, Jul 09 '19 at 16:35

score 2 · Accepted Answer · answered Jul 09 '19 at 16:41

This is not exactly how I would approach the problem here. In particular, returning a lambda is discouraged. Consider instead

import math
import numpy

def aequi_ang_step(step):
    def local_it(start, d_alpha):
        for alpha in numpy.arange(start, start+d_alpha,(d_alpha < 0) and -step or step):
            if (alpha < 2*math.pi): yield alpha
            else: yield alpha-2*math.pi
    return local_it

Now what you return is a closure which includes a definition for step.

I think a more pythonic approach in general for currying, is functools.partial

from functools import partial
import math
import numpy

step = 0.1

def aequi_ang(start, d_alpha, step):
    for alpha in numpy.arange(start, start+d_alpha,(d_alpha < 0) and -step or step):
        if (alpha < 2*math.pi): yield alpha
        else: yield alpha-2*math.pi

aequi_ang_step = partial(aequi_ang, step=step)

Thanks - is there a rationale behind the "don't return lambdas" rule? — Vroomfondel, Jul 11 '19 at 08:30
I'm not sure I know place where that is suggested explicitly, but it seems like an extension of the PEP8 suggestion of not assigning a `lambda` to an identifier. https://www.python.org/dev/peps/pep-0008/#programming-recommendations . When you use a `def` statement instead of a lambda you return an object that is generally more useful and nice since, in particular, it has a name and a better string representation. — Kyle Parsons, Jul 11 '19 at 13:00

Currying iterators in Python

1 Answers1