random tests will not generate words of 0 length now.

Note: there's an additional space somewhere in st that is lost in markdown here.

I agree that the kata is currently incorrect in python. Specifically, test gen produces length-0 words and the reference solution treats length 0 words as words that get sorted first.
It's kind of an interesting test case to have additional spaces/empty words, but as the description doesn't mention it I believe that the test gen is wrong to produce these and should instead produce words of at least length 1 with a single space between each. It's also common for words to ignore empty words, so it is strange to pick this particular definition of word.

it is, but 0 can be counted as False