populating a structured array throws "ndarray is not C-contiguous" error

Question

I am working with a structured array defined in this way:

scores = np.empty((num_of_imgs, 4),
                               dtype=[('id', 'u4'), ('bestT', 'u8'), ('bestR', 'f8'), ('bestP', 'f8')])

then in a for loop I populate it:

scores[i] = [id, bestT, bestR, bestP]

where all the vars inside the list are numpy arrays with shape (1,). However this line of code throws the aforementioned error. Why?

Have you tried "scores[i,:] = [id, bestT, bestR, bestP]"? – Balzola Dec 06 '16 at 11:28 — Balzola, Dec 06 '16 at 11:28

score 4 · Accepted Answer · answered Dec 06 '16 at 11:59

Your scores assignment is making a nx4 array of 4-tuples, which is an extra dimension bigger than you want, I think. It should be
```
scores = np.empty(num_of_imgs,
    dtype=[('id', 'u4'), ('bestT', 'u8'), ('bestR', 'f8'), ('bestP', 'f8')])
```
Then you're trying to assign a list to a tuple, which is throwing your c-contiguous error (numpy isn't as helpful at converting types for structured arrays as it is for ndarrays). Make the assignment a tuple. (using () instead of [])
```
scores[i] = (id, bestT, bestR, bestP)
```

1 Answers1