A toolkit to help develop asynchronous graders for Coursera based on docker images.
This command-line tool is a software development toolkit to help instructional teams author asynchronous graders for Coursera (typically programming assignments). Coursera’s asynchronous grading environment is based upon docker. While use of this tool is by no means required to develop the docker container images, we believe it is helpful in the endeavour. See below for brief descriptions of this tool’s capabilities.
To install this sdk, simply execute:
sudo pip install courseraprogramming
pip is a python package manager. If you do not have pip installed on your machine, please follow the installation instructions for your platform found at: https://pip.pypa.io/en/latest/installing.html#install-or-upgrade-pip
The tool includes its own usage information and documentation. Simply run:
for a complete list of features, flags, and documentation.
Note: the tool requires docker to already be installed on your machine. Please see the docker installation instructions for further information.
Runs a number of sanity checks on your development environment and the Dockerfile that builds your grader to help catch pitfalls early.
These subcommands help you verify that a built docker container image actually has what you expect inside of it. You can use these commands to poke at the file system and verify that everything is where it should be.
Allows for interactive inspection of your docker grading container image to help debug grader issues. By default, it provides a shell that runs in a simulation of the hardened sandbox environment.
This grade subcommand loosely replicates the production grading environment on your local machine, including applying CPU and memory limits, running as the correct user id, mounting the external file systems correctly, and relinquishing the appropriate extra linux capabilities. Note that because the GrID system has adopted a defense-in-depth or layered defensive posture, not all layers of the production environment can be faithfully replicated locally.
The grade subcommand has 2 sub-sub-commands. local runs a local grader container image on a sample submission found on the local file system. The future remote sub-sub-command will run a local grader container image on a sample submission downloaded from Coursera.org. This sub-sub-command is intended to help instructional teams verify new versions of their graders correctly handle problematic submissions.
Allows an instructional team to upload their containers to Coursera without
using a web browser. It is designed to even work in an unattended fashion (i.e.
from a jenkins job). In order to make the upload command work from a Jenkins
automated build machine, simply copy the
~/.coursera directory from a working
machine, and install it in the jenkins home folder. Beware that the oauth2_cache
file within that directory contains a refresh token for the user who authorized
themselves. This refresh token should be treated as if it were a password and
not shared or otherwise disclosed!
To find the course id, item id, and part id, first go to the web authoring
interface for your programming assignment. There, the URL will be of the form:
/:courseSlug/author/outline/programming/:itemId/. The part id will be
displayed in the authoring user interface for each part. To convert the
courseSlug into a courseId, you can take advantage of our
Course API putting in the appropriate courseSlug. For example, given a
course slug of developer-iot, you can query the course id by making the
The response will be a JSON object containing an id field with the value:
This command can also be used to customize the resources that will be allocated to your grader when it grades learner submissions. The CPU, memory limit and timeout are all customizable.
- --grader-cpu takes a value of 1, 2, 3 or 4, representing the number of cores the grader will have access to when grading. The default is 1.
- --grader-memory-limit takes a value of 1024, 2048, 3072 or 4096, representing the amount of memory in MB the grader will have access to when grading. The default is 1024.
- --grading-timeout takes a value between 300 and 1800, representing the amount of time the grader is allowed to run before it times out. Note this value is counted from the moment the grader starts execution and does not include the time it takes Coursera to schedule the grader. The default value is 1200.
Allows an instructional team to publish changes made to programming assignments. Beware that all changes made to your assignment will be published, not just grader changes. Like upload, it is designed to work in an unattended fashion. Multiple items can be published at the same time using the --additional-items flag. There are multiple different error conditions that are represented by exit codes. An exit code of 1 represents a fatal error while an exit code of 2 represents a retryable error.
Please us the github issue tracker to document any bugs or other issues you encounter while using this tool.
Note: We do not have the bandwidth to officially support this tool on windows. That said, patches to add / maintain windows support are welcome!
We recommend developing courseraprogramming within a python virtualenv. To get your environment set up properly, do the following:
virtualenv venv source venv/bin/activate python setup.py develop pip install -r test_requirements.txt
To run tests, simply run: nosetests, or tox.
Code should conform to pep8 style requirements. To check, simply run:
pep8 courseraprogramming tests