ndarray.transpose 和 numpy.transpose 的区别

Question

What is the difference between ndarray.transpose and numpy.transpose ? and in which scenario, we have to which one ? is there a scenario where only one of the above will work ?

I have gone through the documentation and per my understanding numpy.transpose will return a view whenever possible. Whereas ndarray.transpose returns a view always.

It would be great if someone could give me some example where only one of the above is a good fit.

Answer 1

The code for np.transpose is:

def transpose(a, axes=None):
    return _wrapfunc(a, 'transpose', axes)

which is, effectively:

np.asarray(a).transpose(axes)

that is, make a no-copy array and apply the method.

If a is already an array, the two approaches are essentially the same. Either way transpose is a low cost operation, just changing shape and strides attributes. Use which ever makes your code clearest (that is, readable to a human).