TypeError: 'numpy.float64' object does not support item assignment - Similar code, error raises

Question

I am writing a Molecular Dynamics code and for that I have a function that computes forces between particles: conservative, random and dissipative forces. The conservative forces are pairwise forces, which means I have a double loop for to compute them. I wanted to save some time and include the calculation of the random and dissipative forces in one of the loops of the double loop as follows:

fr = np.zeros((npart, dim))
fd = np.zeros((npart, dim))
fc = np.zeros((npart, dim))

for i in range(npart-1):

    for d in range(dim):
        # dissipative and random forces
        fd[i, d] = -gamma * v[i, d]
        fr[i, d] = noise/np.sqrt(dt) * np.random.normal()

    for j in range(i+1, npart):

        # conservative force for particle i
        fc[i, 0] = fc[i, 0] + (dX/r2) * fr
        fc[i, 1] = fc[i, 1] + (dY/r2) * fr
        fc[i, 2] = fc[i, 2] + (dZ/r2) * fr

        # conservative force for particle j (action-reaction)
        fc[j, 0] = fc[j, 0] - (dX/r2) * fr
        fc[j, 1] = fc[j, 1] - (dY/r2) * fr
        fc[j, 2] = fc[j, 2] - (dZ/r2) * fr

Here gamma, noise and dt are constants. I get the following error:

    fr[i, d] = noise/np.sqrt(dt)*np.random.normal()
TypeError: 'numpy.float64' object does not support item assignment

Nevertheless, if I compute the random and dissipative forces in an external, separate loop, the error disappears:

for i in range(npart):
    for d in range(dim):
        fd[i, d] = -gamma * v[i, d]
        fr[i, d] = noise/np.sqrt(dt) * np.random.normal()

What is the difference between both computations? Why there is no error when the computation is done in a separate loop?

your first for loops have different lengths. could it be the source of the problem? — Behzad Mehrtash
– Behzad Mehrtash, Commented Jul 4, 2019 at 8:37
How do you define fr and fc at the beginning? And what of these objects are numpy objects? — micric
– micric, Commented Jul 4, 2019 at 8:41
@micric Sorry, I forgot to add that fr, fc and fd are numpy arrays. I just corrected it now. Only fr, fc and fd, together with v are numpy arrays. — ma7642
– ma7642, Commented Jul 4, 2019 at 9:52
@BehzadMehrtash I printed the outcome of the loops to see what was not working, and the first for loop in 'dim' runs correctly once, then it raises the error. That is, it d = 0, d = 1 and d = 2 without problems for i = 0, and then it raises the error. — ma7642
– ma7642, Commented Jul 4, 2019 at 9:55
fc[i, 0] = fc[i, 0] + (dX/r2) * fr in this line there is a size mismatch, isn't there? fr and fc are the same size at the beginning, but now you are summing fr to just one column — micric
– micric, Commented Jul 4, 2019 at 10:01

ma7642 · Accepted Answer · 2019-07-04 12:39:45Z

1

SOLVED: As @micric pointed out, there is a variable inside the second loop called 'fr', which is of type float. I made the mistake of using the same name for an array. Hence Python's complaints.

answered Jul 4, 2019 at 12:39

ma7642

1492 silver badges12 bronze badges

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

Collectives™ on Stack Overflow

TypeError: 'numpy.float64' object does not support item assignment - Similar code, error raises

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related