dodona-edu / judge-turtle Goto Github PK

When submitting solutions for the Turtle judge, I can't seem to get the images to work:

Tested in both Firefox and Chrome.

(Not entirely sure if this is a turtle judge issue or a Dodona issue, but I'll move the issue to the correct repo if necessary)

• https://naos.ugent.be/nl/submissions/983308/

# import package
import turtle

turtle.shape("turtle")

print("hallo")  # ignore this

# forward turtle by 50
turtle.forward(50)

# print the turtle shape
turtle.stamp()

turtle.speed(20)

# and then continue
# forward turtle by 50
turtle.forward(50)
turtle.bye()

• https://github.com/koalaman/shellcheck/wiki/SC2164
• https://github.com/koalaman/shellcheck/wiki/SC2046
• https://github.com/koalaman/shellcheck/wiki/SC2086
• https://github.com/koalaman/shellcheck/wiki/SC2068

See 5bba9de

Some ideas for visual feedback:

• Bounding box of region of interest with different pixels
• Three viewing modes (with toggle for switching) for comparing submission and solution:
  • Side by side (default)
  • On top of each other with 50% transparency with CSS (and bounding box)
  • Diff of two images (with Pixelmatch or PIL)

Docs:

• https://github.com/whtsky/pixelmatch-py, show difference and number of different pixels
• https://stackoverflow.com/a/196882, calculate difference with PIL and draw bounding box
• https://stackoverflow.com/a/11818358, compare images
• https://github.ugent.be/brewille/turtle/blob/master/compare.py, bounding box with dashed lines
• Interesting way to compare images on GitHub:
  
  

• https://naos.ugent.be/nl/submissions/983304/

# import package
import turtle
import time

turtle.shape("turtle")

# forward turtle by 50
turtle.forward(50)

# print the turtle shape
turtle.stamp()

time.sleep(20)  # ignore this

# and then continue
# forward turtle by 50
turtle.forward(50)
turtle.stamp()

Show error instead.

Edit:

• https://naos.ugent.be/nl/submissions/983425/
  

Python script for teacher which converts turtle solution to image.

Teachers need a png image for the exercise description. This script could be in their GitHub repository with Dodona exercises in a preparation folder. Running the script locally would generate the image in the media folder based on the solution.py.

• Automatic filename based on folder/exercise name
• Option to change canvas size
• svglib? Support for svg's is sufficient, so no png's needed.
• Also write svg image (with option for transparant or white background)

Similar to donkirkby/svg-turtle#4 and based on https://github.ugent.be/brewille/oef_turtle/blob/master/centered_rectangle/preparation/conv2png.py

Show error instead.

• https://naos.ugent.be/nl/submissions/983345/

This feature request tries to make the feedback more meaningful. 2209/2500 pixels correct doesn't say that much, but 88% is already better.

Background pixels are correct by default, so count them separately (see below). E.g. square with a black border and transparant inside. A blank canvas is already 90%+ correct. If the number of black pixels is counted, this results in 0% correct.

• Absolute and percentage difference of total pixels, e.g. 950/1000 (95%) pixels correct (current situation, but add percentage)
• Absolute and percentage difference of mismatched (non-background) pixels, e.g. 900/1000 (90%) foreground pixels correct
• User specifies background color in config.json (default if not mentioned is transparant). e.g. "background-color": "white"

Docs:

• https://github.ugent.be/brewille/turtle/blob/master/non_background.py

Use https://pypi.org/project/colour/ or the color map in svg_turtle to convert between colour formats.

Edit: This is what I meant. The score isn't 90%+, but a more realistic score, 20% similarity.

The default size is ok, if it's not mentioned in config.json.

• https://naos.ugent.be/nl/submissions/983376/

• https://naos.ugent.be/nl/submissions/983293/
• https://github.ugent.be/brewille/oef_turtle/tree/master/circle

Local output:

• https://naos.ugent.be/nl/submissions/983290/
• https://github.ugent.be/brewille/oef_turtle/tree/master/star

This should be the output

Edit: This is apparently because of the custom turtle object. How difficult is this to support? If it isn't worth the hassle we can close it.

• https://github.ugent.be/brewille/oef_turtle/tree/master/custom_canvas
• https://naos.ugent.be/nl/submissions/983379/
• https://naos.ugent.be/nl/submissions/983378/
• https://naos.ugent.be/nl/submissions/983377/

import turtle

center = (0, 0)
side = 50

turtle.fillcolor("blue")
x_pt, y_pt = center

x_pt -= side / 2
y_pt += side / 2

turtle.up()
turtle.goto(x_pt, y_pt)
turtle.down()

turtle.begin_fill()
for _ in range(4):
    turtle.forward(side)
    turtle.right(90)
turtle.end_fill()

turtle.home()
turtle.dot(5)


turtle.done()

Standard Error:
Traceback (most recent call last):
  File "/mnt/_n9h4GBuCwjJb87ulQw3cg/judge/turtle_judge.py", line 74, in <module>
    correct_pixels, total_pixels = diff_images(png_submission, png_solution)
  File "/mnt/_n9h4GBuCwjJb87ulQw3cg/judge/judge/runtime.py", line 61, in diff_images
    diff = ImageChops.difference(image1, image2)
  File "/usr/local/lib/python3.10/site-packages/PIL/ImageChops.py", line 102, in difference
    return image1._new(image1.im.chop_difference(image2.im))
ValueError: images do not match

Load custom backgrounds in turtle canvas with bgpic.

Edit: This feature request is to make "connect-the-dots"-exercises possible. This would be ideal to learn about the coordinate system and get some visual help. E.g.