Staffeli NT Technology
With Staffeli, we work with local course clones. We aim to keep these clones compatible with git.
We recommend that you create a local directory canvas
,
absalon
, or similar, for all of you Canvas-related local course
clones.
Staffeli needs some initial help to be able to login with your
credentials. You need to generate a
token
for Staffeli to use, and save it in your home directory in a file with
the name .canvas.token
.
NB! This is your personal token so do not share it with others, else they can easily impersonate you using a tool like Staffeli. Unfortunately, to the best of our knowledge, Canvas has no means to segregate or specialize tokens, so this is really "all or nothing".
$ pip3 install -r requirements.txt
Or you can install in a virtual environment:
-
Create a virtual environment called
env
.On macOS and Linux:
$ python3 -m venv env
On Windows:
$ py -m venv env
-
Activate
env
On macOS and Linux:
$ source env/bin/activate
On Windows:
$ .\env\Scripts\activate
-
Now install the requirements for
staffeli_nt
inenv
$ pip3 install -r requirements.txt
There are multiple options for fetching submissions.
The general command is python <staffeli_nt_path> download <course_id> <template.yaml> <assignment-dir> [flags]
, where
<staffeli_nt_path>
is the path to the directory wherestaffeli_nt
is located, i.e. where the filesdownload.py
andupload.py
etc. can be found.<course_id>
is the canvascourse_id
for the course.<template.yaml>
is the template file to use when generating thegrade.yml
file for each submission<assignment_dir>
is a non-existing directory, that staffeli will create and store the submissions in.
Windows:
Since staffeli_nt
is written in python3
, you will need to invoke it via your python3
interpreter.
Example: python <staffeli_nt_path> download <course_id> <template.yml> <assignment-dir> [flags]
Fetching all submissions:
To fetch all submissions from the course with id 12345
, using the template-file ass1-template.yml
and create a new directory "ass1dir" to store the submissions in:
$ python <staffeli_nt_path> download 12345 ass1-template.yml ass1dir
This will present you with a list of assignments for the course, where you will interactively choose which assignment to fetch.
For each submission, a directory will be created in <assignment_dir>
, in which the handed-in files of the submission will be stored, alongside a file grade.yml
generated form the <template.yml>
for a TA to fill out during grading of the assignment.
Submission comments, if any, will be downloaded as well, and stored alongside grade.yml
and the files of the hand-in.
In case the student hands in a file called grade.yml
it will be overwritten by staffeli. If the student hands in a file called submission_comments.txt
and has written submission comments on the submission, the downloaded submission comments will be renamed.
What we call "Hold", canvas/absalon calls sections.
To fetch all submissions for an assignment, where the student belongs to a given section, and the <course_id>
is 12345
:
$ python <staffeli_nt_path> download 12345 ass1-template.yml ass1dir --select-section
This will present you with a list of assignments for the course, where you will interactively choose which assignment to fetch, followed by a list of sections for you to choose from.
It is possible to fetch specific submissions based on a list of kuids. To do this, create a YAML-file with the following format:
TA1:
- kuid1
- kuid2
- kuid3
TA2:
- kuid4
- kuid5
To then fetch all submissions for an assignment for a given TA:
$ python <staffeli_nt_path> download <course_id> ass1-template.yml ass1dir --select-ta ta_list.yml
where ta_list.yml
is a YAML-file following the above format.
This will present you with a list of assignments for the course, where you will interactively choose which assignment to fetch, followed by the list of TA's from your ta_list.yml
file.
Selecting a TA, will fetch submissions from each kuid
in the file, associated with the chosen TA, i.e. selecting TA1
will fetch submission from kuid1
, kuid2
and kuid3
.
In the template.yml
-file you can add a field:
onlineTA: https://address.of.onlineTA.dk/grade/assignmentName
This will (attempt to) run onlineTA for each downloaded submission.
It is possile to only fetch submissions that are either ungraded or have a score < 1.0.
Currently this is implemented specifically for the PoP-course and might not be available in the current form in later releases.
This can be achieved by appending the --resub
flag to any use of the download
subcommand.
Use python <staffeli_nt_path> upload <template.yaml> <assignment-dir> [--live] [--step]
.
The default to do a dry run, that is not to upload anything
unless the --live
flag is given.
For instance, to review all feedback for submissions in the directory
ass1
before uploading:
$ python <staffeli_nt_path> upload ass1-template.yml ass1 --step
To upload all feedback for submissions in the directory
ass1
:
$ python <staffeli_nt_path> upload ass1-template.yml ass1 --live
To upload feedback for a single submission:
$ python <staffeli_nt_path> upload-single <POINTS> <meta.yml> <grade.yml> <feedback.txt> [--live]
To generate feedback.txt
locally for submissions in the directory ass1
:
$ python <staffeli_nt_path> upload ass1-template.yml ass1 --write-local
A (minimal) template could look like:
name: Mini assignment
tasks:
- overall:
title: Overall
points: 6
rubric: |
Some default feedback.
Your code and report are unreadable.
Wow, that's really clever.
The template files support a few optional fields.
-
passing-points: N
: Adding this field will have the effect, that the grade posted is1
if the total sum of points is greater than or equal topassing-points
, and0
otherwise. -
show-points: BOOL
Setting show-points tofalse
will exclude thepoints/grade
from the generatedfeedback.txt
files. Use this, if you do not want the students to see the points-per-task, but only receive an overall grade. -
onlineTA: ADDR
Include this field to (attempt to) run onlineTA at addressADDR
for each submission, when downloading submissions.
name: Mega assignment
passing-points: 42
show-points: false
onlineTA: https://yeah-this-is-not-a-real-address.dk/grade/megaassignment
tasks:
- megaAssignmentGeneral:
title: Mega assignment - General comments and adherence to hand-in format requirements
points: 100
rubric: |
[*] You should spell check your assignments before handing them in
[-] You are using the charset iso-8859-1. Please move to the modern age.
[-] Your zip-file contains a lot of junk. Please be aware of what you hand in.
- megaAssignmentTask1:
title: Task 1
points: 2
rubric: |
[+] Your implementation follows the API
[-] Your implementation does not follow the API
[+] Your tests are brilliant
[-] Your tests are not tests, just print-statements.
This is equivalent to an exam without an examinator, where you shout
in a room for half an hour and give yourself the grade 12.
- megaAssignmentTask2:
title: Task 2
points: 2
rubric: |
[+] Very good points.
[+] Very good points. However, I disagree with ...
[-] I fail to comprehend you answer to this task.
- megaAssignmentBonusTask:
title: Bonus tasks that do not give points, or another option for general comments
rubric: |
[*] You did extra work! It won't help you though.