nbgrader cell grading issues: kernel not agnostic, confusing, English

Currently, when you create a nbgrader notebook and mark a cell to be autograded, it will replace the cell’s contents (if any) with:

# YOUR CODE HERE
raise NotImplementedError()

This is confusing for two reasons:

1. If you actually put in code, such as:

(define my-test-function
    (lambda (arg)
        ;; your code here...

it is just removed.

1. It is replaced with Python code. Beginning Python students are confused by these lines. It is very confusing if they are beginning students of a different language.

Constraints: there is an automatic feature that will replace/search for text, such as :

### BEGIN SOLUTION
...
### END SOLUTION

This is, of course, language specific (eg, the comment hash is specific to Python).

A related problem is that text Markdown cells that are marked for autograding are also removed. For example, the text is always replaced with:

YOUR ANSWER HERE

My overall feeling is that nbgrader is attempting to do too much, and that causes issues because what it does do is not kernel (nor human-language) agnostic. If it instead left the text/code alone the cells, many of these issues would go away.

But what to do about automatic solution hiding? Perhaps something more general:

• if a line contains the text “*** begin solution ***” (can be embedded in comments of kernel), then start of solution
• if a line contains the text “*** end solution ***” (can be embedded in comments of kernel), then end of solution

Also, the system should not rely on raising NotImplementedError(). This is confusing to see, and is a weak way to check for “validation”.

[willingc]

@jhamrick @dsblank Happy to help on this and other 0.4.0 milestones. You know my heart is with education and Jupyter/JupyerHub, and Doug is doing very cool stuff. I’ll give the notebook template a closer look tomorrow.

P.S. Nice to have you back @jhamrick.

[jhamrick]

To summarize, a few issues came out of this discussion:

• #543 – document how to change the default code and text stubs to be non-Python and non-English
• #547 – indicate in the create assignment UI what will be overwritten when nbgrader assign is run
• #549 – give better feedback to students about how many points they have earned
• #551 – be more flexible in finding solution delimeters

I think the bulk of what we talked about in this issue has been turned into these other issues, so I am going to close this. @dsblank if you disagree, please let me know and either I or you can open up new issues for specific things that should be addressed.