When trying to retrieve several labels and it fails, it is sometimes hard to know exactly which label(s) are missing from the array, especially when the axes are long and not all labels are displayed in the axes summary.

I think that when it fails, displaying axes where it matches partially and the missing keys could be a large quality of life improvement:

>>> arr = la.ndtest('a=a0..a4,a6..a9;b=b0..b2')
>>> arr['a0,a7,a2,a5,a3,a4,a8']
ValueError: 'a0,a7,a2,a5,a3,a4,a8' is not a valid label for any axis:
 a [9]: 'a0' 'a1' 'a2' ... 'a7' 'a8' 'a9'
 b [3]: 'b0' 'b1' 'b2'

FWIW, as a quick workaround, I had to resort to this to know which key was missing:

>>> arr.a.union('a0,a7,a2,a5,a3,a4,a8').difference(arr.a)
Axis(['a5'], 'a')