make scalar python function accept list and arrays

Question

In python, is there an easy and efficient way to make function f(x: float) accept both lists and numpy arrays as arguments (in which case I would want to apply f element-wise and return the result in the same format as it was sent in)? For now, I need only 1-dimensional arrays.

As an illustration (my real f is more complex), let's say that I have:

def f(x):
    return math.log(x) if x > 0 else 0.0

Then this one works, but is not that elegant - and possibly not that efficient either, because of the recursion (which I use as it allows me to have just one function):

def f(x):
    if np.isscalar(x):
        return math.log(x) if x > 0 else 0.0
    elif isinstance(x, np.ndarray):
        return np.array([f(i) for i in x], dtype=float)
    else:
        return [f(i) for i in x]

Is there a better way?

@hilberts_drinking_problem: but this would create new function that returns an numpy array. Since I want the output in the same format as input (scalar/list/array), this would need at least two functions and still some ifs to decide what to do - how is it better than my version? — Michal Kaut
– Michal Kaut, Commented Apr 15, 2020 at 9:19
Preserving the input type sounds like a single/multiple dispatch issue, which is a big deal in language design. You may be able to simplify the code for performing this in Python, but I doubt if you can gain much performance as of today. — hilberts_drinking_problem
– hilberts_drinking_problem, Commented Apr 15, 2020 at 9:54
You could probably use a decorator for a clean pythonic way to add pre- and post- call processing to your function. — RufusVS
– RufusVS, Commented Feb 27, 2021 at 22:11

ulmo · Accepted Answer · 2021-02-27 22:44:43Z

3

Using a decorator function would be a good option if this is something you'd need to for more than just a single function definition in your code-base. It does mean that you have to wrap your code, but I'm assuming that what you mean by "having just one function" is that you do want the function expression in f(x) to be the same for all cases.

Using your existing code, the decorator function would look like this:

def elementwise_possible(func):
    def wrapper(x):
        if np.isscalar(x):
            return func(x)
        elif isinstance(x, np.ndarray):
            return np.array([func(i) for i in x], dtype=float)
        else:
            return [func(i) for i in x]
    return wrapper

And you would write your function like this:

@elementwise_possible
def f(x):
    return math.log(x) if x > 0 else 0.0

The resulting output becomes

In[2]: A = f(2)
In[3]: A
Out[3]: 0.6931471805599453

In[4]: B = f(np.array([2,3,4]))
In[5]: B
Out[5]: array([0.69314718, 1.09861229, 1.38629436])

In[6]:C = f([5,6,7])
In[7]:C
Out[7]: [1.6094379124341003, 1.791759469228055, 1.9459101490553132]

I think efficiency should be the same.

edited Feb 27, 2021 at 22:44

answered Feb 27, 2021 at 22:08

ulmo

313 bronze badges

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

1 Comment

RufusVS Over a year ago

I should have read your answer before adding my comment to the question. Good one!

Collectives™ on Stack Overflow

make scalar python function accept list and arrays

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related