02-numpy: clarify numpy.diff() part

[elcorto]

Using a variable called npdiff to hold an array is highly confusing,
given that the function to be learned here is called {numpy,np}.diff().

Also, stress that the array returned by diff() is shorter by one.

[maxim-belkin]

Thanks, Steve. I agree with your reasoning and like the changes you make to the text. I wonder, though, if a is a good name for a variable. Shall we use a full word like array or something else? My co-maintainers, Anne and Lauren, might have some ideas.

Anyway, thanks for the pull request!

[ldko]

Hi Steve, I agree with Maxim--I like the clarifications your edits contribute to the text but would prefer to avoid the short variable name a and try for something slightly more meaningful (realizing we are giving a generic example which makes it difficult to be very specific). I would, however, also prefer to avoid calling the variable array because of numpy having an array function and the output in the exercise appearing as array([2, 3, 4, 5]) which could confuse some learners into making a connection between the variable name array and the function. I would be okay with something like values or data or value_array etc.

[elcorto]

Valid points regarding a as variable name. I was thinking in numpy docstring terms where short variable names are often used, but of course in this context another naming is needed. I just pushed an update.

[maxim-belkin]

Thanks, Steve. I had the same concern as Lauren so I'm glad you reverted that change.

In any case, a question to both of you: what if we name our variable patient3_week1 (or something like that)? I know this is not the best name for a variable but... it makes it easy to understand what information the variable contains.

[elcorto]

I didn't do that since the original example uses an artificial 1d array of length 5 (enough for explaining what diff() does) instead of pulling a row (length 40) from the data array read in earlier in the episode, probably by design. My compromise was adding row_start = data[i,:5] in the text to make a connection from the following 1d example to the 2d array data, which wasn't present before.

[ldko]

I am ok with using row_start, since the data doesn't actually map to a patient but hints at the possibility.

[maxim-belkin]

row_start is OK... What I was trying to say is that this part of the lesson is currently independent of the lesson and I suggest to blend it in a little bit. For example, we can say that to see if there are any peculiarities in the behavior of inflammation of a particular patient over time, we can use numpy.diff() function. But we can certainly do it in a separate PR.

[ldko]

If @elcorto wants to modify the exercise to reflect the actual patient data and further blend the exercise into the lesson as you suggest @maxim-belkin, I think that would be great, otherwise, I think it is ok to merge as-is and further blend it in a separate PR.

[maxim-belkin]

Agreed.

[elcorto]

Unfortunately, I won't have time to improve this episode more for now in the way you outlined. However, I found other issues that I hope to fix in later PRs.

As for this PR, I don't have any pending changes to add.