ValueError: setting an array element with a sequence

Question

These two sections my code are giving the following error message.

def Simpsons(a, b, n, f, x):
    h=(b-a)/n
    s = 0.
    step = a + h 
    for i in range(1, int(n/2)):
        s += 4*f(x)
        #step += 2*h

    step = a + 2 * h
    for i in range(2, int(n/2 - 1)):
        s += 2*f(x)
        #step += 2 * h

    area = (h/3) * (f(a) + f(b) + s)
    return area

and

for i in range(0, len(list_n)):
    for j in range(0, len(Functions)):
        x = np.linspace(a, b, list_n[i])
        error_simp[j, i] = Exact_intergrals[j] - Simpsons(0, 1,         list_n[i], Functions[j], x)
print(error_simp)

Give this message:

   error_simp[j, i] = Exact_intergrals[j] - Simpsons(0, 1, list_n[i], Functions[j], x)
ValueError: setting an array element with a sequence.

Why is this? everything I have tried does not get rid of it.

Make sure if you post Python code that you reproduce your indentation accurately. Otherwise you're introducing new problems in the code people are trying to find problems in. — khelwood, Dec 02 '16 at 11:41
side remark: you should make a case destinction in Simpson (integration) if n is even or odd. The simpson rule can be extended to an odd number of intervalls, by handling the last intervall seperately (Fitting a parabola into the last 3 points and integrate between the last 2 points only) — Markus Dutschke, Dec 02 '16 at 12:16
What is `error_simp`? Shape, dtype? What does the RHS produce? — hpaulj, Dec 02 '16 at 12:45

score 0 · Answer 1 · edited May 23 '17 at 12:30

I guess you are trying to load array into error_simp[j,i] which might be able to contain only float numbers.

Here is a simple example which illustrates this error:

import numpy
b = numpy.zeros(5)
b[0] = numpy.zeros(5) - 1

Produces the same error.

Looks like either Simpsons function returns an array or Exact_intergrals[j] contains an array.

If you need to save arrays in error_simp matrix, try to add one more dimension:

error_simp = np.zeros((rows, cols, additional_dimension_size))

(Another option is to explicitly set dtype=object option in constructor, but it's much slower and not recommended).

There are a few similar questions that might help: this one and this

score 0 · Answer 2 · answered Dec 02 '16 at 19:23

Your Simpsons function when given an array returns an array of the same size:

In [335]: Simpsons(0,1,10, lambda x:x**2, np.linspace(0,1,3))
Out[335]: array([ 0.03333333,  0.2       ,  0.7       ])

You don't show how error_simp is initialized. The use error_simp[j, i]=... and the error suggests that

np.zeros((n,m), float)

or something like that.

For further help you need to specify what you expect to put in error_simp. If x is the same for all elements, you could create a matching 3d array. If x has different lengths, it will have to be dtype=object. But I like to discourage this usage by beginners. Such an array is little more than a glorified (or debased) list of lists. It's easy to have unrealistic expectations of what you can do with such an array.

=========

Looking more carefully at

for i in range(0, len(list_n)):
    for j in range(0, len(Functions)):
        x = np.linspace(a, b, list_n[i])

you probably are creating different size x. You don't show list_n, but it probably is something like [10,12,14,...], testing this function for different n.

ValueError: setting an array element with a sequence

2 Answers2