Autograde runs manually graded answers

[makhalin]

Operating system

Ubuntu 16.04.3 LTS

nbgrader --version

Python version 3.6.2 | packaged by conda-forge | (default, Jul 23 2017, 22:59:30)
[GCC 4.8.2 20140120 (Red Hat 4.8.2-15)]
nbgrader version 0.5.3

jupyterhub --version (if used with JupyterHub)

0.7.2

jupyter notebook --version

5.0.0

Expected behavior

nbgrader autograde doesn't run "Manually graded answer"

Actual behavior

nbgrader autograde runs "Manually graded answer"

Steps to reproduce the behavior

1. Create cell with "Manually graded answer".
   
2. Generate -> Realease -> Collect
3. Run autograde on one notebook.

I only want to run autograde cells. bacause manual cells have very "heavy" code and I've got time limit on this cells.

[jhamrick]

Hmm, I can see the usecase for this though it sort of goes against the notebook philosophy that the notebook should be runnable beginning to end (and if some cells in the middle are skipped, this might not be the case---for example if a later cell uses the A variable that you've defined in the manually graded answer cell if your example).

Would an acceptable compromise be to allow specific notebooks to be skipped from execution, rather than specific cells? (Note that you can currently disable execution for the entire assignment with --no-execute, but I assume this is not what you want).

[makhalin]

My notebook has both type of cells.

[jhamrick]

I figured that probably was the case. Would it be possible to refactor the problem into two notebooks, e.g. part a and part b, one of which is manually graded and one which is auto graded (or at least has no execution skipping)?

[makhalin]

I'll try it. But can you add this kind of cell's type in future versions?