unbounded/less bounded size filters for Object Classification from Pixel Prediction Map about ilastik HOT 2 CLOSED

The commit that introduced the 100000 max doesn't mention any reason. I'm guessing this was just a good-enough value chosen to avoid having to implement logic coupling one spinner's limit to the other spinner's value (or any logic, for that matter...)

Choosing a min > max doesn't actually seem to cause any problems (besides obviously leading to nothing passing the filter), so we could just remove the max on the spinner altogether...

Looks like it's only a GUI limit btw, so one could still choose any value by manually tweaking the .ilp 😁

from ilastik.

I think I found out why the numbers were chosen... I tried removing the maximum altogether:

Now Qt doesn't allow values above 99 🙃 It's also clear that the maximum controls the width of the spinner input 🤦 This is the result with maxima approaching the highest numbers an int can hold:

So the two numbers were probably chosen to get a suggestive width for the box (smaller box for the size filter Min).

This ends up getting quite wide for an applet that already gets a horizontal scroll bar with the default drawer width. So I think I would go with vertical positioning for these two instead:

from ilastik.