Map like function with NumPy and Pandas

The best way is to create a numpy ufunc

You can create your own custom ufuncs in NumPy using the numpy.frompyfunc() function or by using the numpy.vectorize() decorator.

Custom ufuncs can be useful when you have a specific mathematical operation that you want to apply element-wise to arrays. They provide a way to extend NumPy’s functionality with your own custom functions.

1. numpy.frompyfunc:
   • Purpose: numpy.frompyfunc creates a universal function (ufunc) from an ordinary Python function.
   • Parameters:
     • func: The Python function that you want to convert into a ufunc.
     • nin: Number of input arguments expected by the function.
     • nout: Number of output arguments returned by the function.
   • Returns: A ufunc object.
2. numpy.vectorize:
   • Purpose: numpy.vectorize is a class that transforms a Python function taking scalar inputs into a vectorized function.
   • Parameters:
     • pyfunc: The Python function you want to vectorize.
     • otypes: Output data types of the vectorized function. Default is None.
     • doc: Optional docstring for the vectorized function.
     • excluded: Indices or slices of arguments to exclude from vectorization. Default is None.
     • cache: If True, cache the first function call. Default is False.
     • signature: Signature of the vectorized function. Default is None.

Here are examples of creating custom ufuncs using both methods

import numpy as np

# create the function you want to execute
f = lambda x, y: x * y

# alternative 1
f_arr = np.frompyfunc(f, 2, 1)

# alternative 1
vf = np.vectorize(f)

# create the numpy array
arr = np.linspace(0, 1, 10000)

# performance
%timeit f_arr(arr, arr) # 307ms
%timeit vf(arr, arr) # 450ms

Both frompyfunc and vectorize serve similar purposes, but they have some differences in usage and behavior. frompyfunc is generally faster but less flexible than vectorize. Additionally, vectorize allows more control over the output data types and other aspects of the vectorized function.

numpy.frompyfunc and numpy.vectorize are two functions in the NumPy library that allow you to create universal functions (ufuncs) from Python functions. These functions can then be applied element-wise to arrays.

Here’s a brief explanation of each along with the parameters they need: