How to code a function that accepts float, list or numpy.array?

Question

I have the following simple Python function:

def get_lerp_factor( a, x, b ):
    if x <= a: return 0.
    if x >= b: return 1.
    return (x - a) / (b - a)

Many numpy functions, like numpy.sin(x) can handle a float or an array.

So how can I extend this in the same manner, so that it can also handle a numpy array for x?

def get_lerp_factor( a, x_maybe_array, b ):
    out = (x_maybe_array - a) / (b - a) # this should work...
    # but now I have to clamp each element of out between 0 and 1

Would I have to specifically check the type of x, and branch accordingly?

How about:

def get_lerp_factor( a, x_anything, b ):
    x = np.array( x_anything )
    out = ...(x)
    # now typecast out back into the same type as x... will this work?

?

Not a very good idea IMO, it breaks the single responsability principle. — Paulo Bu
– Paulo Bu, Commented Feb 28, 2014 at 12:18
Why not handle this for a single value and then use map or a list comprehension to deal with lists and arrays? — msvalkon
– msvalkon, Commented Feb 28, 2014 at 12:23
@PauloBu: Most NumPy functions behave in this way (taking any arguments that can be converted to arrays) so the single responsibility ship has long since sailed. — Gareth Rees
– Gareth Rees, Commented Feb 28, 2014 at 13:02
@Josay: It should return a NumPy array containing the lerp factor elementwise. — Gareth Rees
– Gareth Rees, Commented Feb 28, 2014 at 13:03

Gareth Rees · Accepted Answer · 2014-02-28 12:51:41Z

19

You need numpy.asarray. This takes as its first argument:

Input data, in any form that can be converted to an array. This includes lists, lists of tuples, tuples, tuples of tuples, tuples of lists and ndarrays.

and it returns:

Array interpretation of a. No copy is performed if the input is already an ndarray.

So you can implement your function like this:

import numpy as np

def get_lerp_factor(a, x, b):
    a, x, b = np.asarray(a), np.asarray(x), np.asarray(b)
    return ((x - a) / (b - a)).clip(0, 1)

This works for scalars:

>>> get_lerp_factor(0, 9, 16)
0.5625

and also for iterables:

>>> get_lerp_factor(2, range(8), 6)
array([ 0.  ,  0.  ,  0.  ,  0.25,  0.5 ,  0.75,  1.  ,  1.  ])

answered Feb 28, 2014 at 12:51

Gareth Rees

66.1k10 gold badges139 silver badges166 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Daren Thomas · Accepted Answer · 2014-02-28 12:28:43Z

Actually, as long as the numpy array has the semantics you want for the operators you're using (-, /, <=, >=), then it already works. This is called "duck typing", where you don't really care what type your arguments have, just that they behave in a specific way.

Of course, a list is not going to behave like that. And the numpy array might not behave entirely like that either (the <= operator works, but results in an array of booleans, the => operator is not defined). So, you will have to check the type at runtime. One way to do this is to check if the __len__ method is supported:

try:
    if len(a) == len(b) == len(x):
         # handle list / array case
except TypeError:
    # oops, was a float

Note that it is generally a very bad idea to check the type explicitly (with isinstance(o, t)) since you want to maintain as much duck type semantics as possible. But sometimes you need to do just that.

Note also, that this is only really acceptable when the "semantics" of the function remain the same, regardless of the input type. If you are changing the meening of the function based on input types, then you're going to have a bad time! Or worse: Your users (consumers of your function) are going to have a bad time.

Collectives™ on Stack Overflow

How to code a function that accepts float, list or numpy.array?

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related