Where are we at with CodeRunner development and new versions?

Where are we at with CodeRunner development and new versions?

por Tim Hunt -
Número de respuestas: 2

Hi Richard, I was giong to send you a private message, but then I thought this might be a useful question for others to see the answer to, and so I will ask it here.

I can see on GitHub that there is interesting work going on on the development branch of qtype_coderunner, and related chanages in the master branch or qbehaviour_adaptive_adapted_for_coderunner. How close is that to release? (We are in no real hurry to do an upgrade.)

Thanks.

En respuesta a Tim Hunt

Re: Where are we at with CodeRunner development and new versions?

por Richard Lobb -
Hi Tim. Thanks for the reminder - it probably is about time for us to merge most of the changes into Master and release.

The most major change is the introduction of a cache between CodeRunner and Jobe to prevent rerunning all the jobs during a regrade. This has been a huge win for us with our class of 1000 students, reducing the time for a full regrade of an assignment or exam from around a day to a few tens of minutes. We've been running that functionality all semester here and I think we can safely release it once we've moved the control variables back into the plugin settings. I'd like to run the release candidate on our production server for a week or two of our second semester, which starts next week. So hopefully we can release all the changes within a month.

I appreciate that the new Moodle functionality to restrict regrading to just selected questions will also dramatically cut back on regrading costs, but the two changes should work really well together.

One change we're holding back is the introduction of the functionality that allows combinator template grader questions to return files (usually images). Currently our matplotlib questions return images as data-URIs embedded in the html. This works fine but is very expensive on disk space, with the same image being stored hundreds or even thousands of times. Also¸ our IT department is unhappy with the use of data-URIs as they're considered a security hole. As you'll know from our earlier private correspondence, my relatively straightforward implementation of the functionality to allow combinator graders to return files that could be stored in the Moodle file system failed when courses were exported with full user data and then reimported - Moodle's hacky way of updating embedded URLs broke the serialisation of the grader outcome. I did implement your suggestion of switching to JSON serialisation throughout instead of the standard PHP serialisation but I'm not happy with it -- too much complexity for functionality that most users won't even know about, let alone use. I'm considering an alternative implementation where CodeRunner images and other files are accessed via a custom web-service but there's no timeline on that. Probably not until our Southern Hemisphere summer break.
En respuesta a Richard Lobb

Re: Where are we at with CodeRunner development and new versions?

por Tim Hunt -
Thanks Richard. I agree with your plan of running the release candidate on our production server for a while.

Therfore, don't rush because of us. We will stick with the current version for the next few months, and plan to take the next version in late October, ready for are December server update. Thanks.