Python numpy array index can be -1?

Question

I'm trying out opencv samples from https://github.com/Itseez/opencv/blob/master/samples/python2/letter_recog.py and I need help deciphering this code..

new_samples = np.zeros((sample_n * self.class_n, var_n+1), np.float32)
new_samples[:,:-1] = np.repeat(samples, self.class_n, axis=0)
new_samples[:,-1] = np.tile(np.arange(self.class_n), sample_n)

I know what np.repeat and np.tile are, but I'm not sure what new_samples[:,:-1] or new_samples[:,-1] are supposed to do, with the -1 index. I know how numpy array indexing works, but have not seen this case. I could not find solutions from searching.

Leb Leb · Accepted Answer · 2015-12-06T22:24:53

Python slicing and numpy slicing are slightly different. But in general -1 in arrays or lists means counting backwards (from last item). It is mentioned in the Information Introduction for strings as:

>>> word = 'Python'
>>> word[-1]  #last character 
'n'

And for lists as:

>>> squares = [1, 4, 9, 16, 25]
>>> squares
[1, 4, 9, 16, 25]
>>> squares[-1]
25

This can be also expanded to numpy array indexing as in your example.

new_samples[:,:-1] means all rows except the last columns

new_samples[:,-1] means all rows and last column only

Python numpy array index can be -1?

1 Answers