Lots of "found contradiction" with these patterns

[delhoume]

I tried many combinations of pattern and output size and only 2 32 32 seems to work with these input images
For example:
python3 waveFunctionCollapse.py flash_invaders8x8.png 3 32 32
|----------------------------------------------------------------------------------------------------| 0%
found contradiction

same with <pattern_size> greater than 2 and larger output sizes.

What can go wrong ?
When it works (2 32 32) it looks quite nice !

Also on my mac I cannot get the <render_iterations> to not pop up the window
WaveFunctionCollapse % python3 waveFunctionCollapse.py flash_invaders10x8.png 2 32 32 True True False
always pops annoying windows endlessly...

[avihuxp]

The "Found contradiction" error occurres when the algorithm reaches a point where for some cell, no patterns are possible.
This current implementation of the code does not have a way to handle that on it's own (a simple solution is to run the algorithm several iterations and hope that one of them will be successful, a more elegant solution will be to implement backtracking (PR's are more than welcome 😬)).
As for what can be done:

• I personally rarely got this issue, so i would try to fidget with the input parameters (mainly the pattern size if i had to guess).
• I would also recommend trying with and without both flipping and rotating the patterns on extractions (which can be done using the optional command line parameters).
• For a more hands on debugging, you can you the function that saves all patterns to the device, which is found in the python file, to check out each and every pattern. This way you can try and match for yourself and see if there is some pattern that does not have any other patterns to fit it, although this would be very unusual.

As for the windows popping up, these are the matplotlib plots... I'm using Windows, but I'll try to check why the windows keep popping up.

Anyway, I will try to take a look at your inputs, who knows, maybe (and somewhat likely 😬) there's a bug in the code🤷🏻‍♂️

[delhoume]

thanks

[OrrMatzkin]

More insights about the option to show the algorithm iterations....

When using plt.show() the default behaviour is to wait for all figures to be terminated before returning, meaning the main loop is halted until all figure windows are closed. A good way to display the iterations while still running the main loop is to use interactive mode.