Creating list of objects using data from numpy arrays

Question

I have several numpy arrays containing data, and a class I defined, with attributes corresponding to each of my numpy arrays. I would like to quickly make another array containing a list of objects, with each objects attributes defined by the corresponding element of the numpy array. Basically, the equivalent of the following:

class test:
 def __init__ (self, x, y): 
 self.x = x
 self.y = y 
a = np.linspace(100)
b = np.linspace(50) 
c = np.empty(len(a), dtype = object) 
for i in range(len(a)):
 c[i] = test(a[i], b[i])

So, my question is does there exist a more concise and pythonic way to do this?

Thanks in advance!

Answer 1

That iteration is 'pythonic'.

numpy does have a function that works in this case, and may be a bit faster. It's not magical:

In [142]: a=np.linspace(0,100,10)
In [143]: b=np.linspace(0,50,10) # change to match a size
In [144]: f=np.frompyfunc(test,2,1)
In [145]: c=f(a,b)
In [146]: c
Out[146]: 
array([<__main__.test object at 0xb21df12c>,
 <__main__.test object at 0xb21dfb2c>,
 <__main__.test object at 0xb221a9cc>,
 <__main__.test object at 0xb222c44c>,
 <__main__.test object at 0xb2213d0c>,
 <__main__.test object at 0xb26bc16c>,
 <__main__.test object at 0xb2215c0c>,
 <__main__.test object at 0xb221598c>,
 <__main__.test object at 0xb21eb2cc>,
 <__main__.test object at 0xb21ebc6c>], dtype=object)
In [147]: c[0].x,c[1].y
Out[147]: (0.0, 5.555555555555555)

frompyfunc returns a function that applies the input f to elements of a,b. I define it as taking 2 inputs, and returning 1 array. By default it returns an object array, which suits your case.

np.vectorize uses this same function, but with some overhead that can make it easier to use.

It also handles broadcasting, so by changing an input into a column array I get a 2d output:

In [148]: c=f(a,b[:,None])
In [149]: c.shape
Out[149]: (10, 10)

But keep in mind that there isn't a lot that you can do with this c. It's little more a list of test instances. For example c+1 does not work, unless you define a __add__ method.