My list of of an array can be appended but my slice can not. I don't understand why.

Hi Stephen,

Yes, I can see where the confusion would come from! In the Teacher's Notes, I think Craig was trying to get at the general fact (not related to the specific example in the video) that there is more than one way to get a 2D array from a 1D array - I don't think he was trying to propose this particular code (with the nd.newaxis) as an equivalent substitute to what was laid out in the video.

The issue with the second line you mention above, with nd.newaxis, is that it does create a 2D array, but it creates it in the opposite direction to what you need. In other words, if I isolate just fake_log[:, np.newaxis], I get:

array([[132],
       [122],
       [128],
       [ 44],
       [136],
       [129],

       ... etc ...

       [110],
       [ 58],
       [ 93],
       [ 79]], dtype=uint16)

And if I do np.shape of that, I get (100, 1), i.e. 100 rows that are all 1 column each. This is the opposite of [fake_log] in the original example, which is (1, 100), or 1 row that has 100 columns.

The error you get when you're trying to append the (100, 1) matrix to the study_minutes, which contains arrays that are all (1, 100) is that the number of rows and columns don't match up like they need to. In theory, you could fix the problem by swapping the np.newaxis with the : (i.e., the rows and columns), like so:

new_fake_log = np.append(study_minutes, fake_log[np.newaxis, :], axis=0)

But [fake_log] is easier and gives the same result, so might as well just do that. :-)

The flattened array when you don't specify an axis is the typical default behaviour for np.append. This would have happened even if the array dimensions had matched, like:

new_fake_log = np.append(study_minutes, [fake_log])

I hope this helps clarify things a bit!