Any way to deal with repeats?

[CamiloMM]

Say I have a wildcard of numbers 1 to 10. Then I get:

• 4
• 7
• 4
• 3
• 4
• 4

It's obvious I did not want this. On a large generation, I might generate a grid of 30 images with a wildcard with 50 entries; I don't need repeats, and I'll probably get 20. Or on a grid of 6, I don't want a repeat. I know this is how randomness works, but it's not what people expect.

Is there any way there could be an "avoid repeats" checkbox or a "max repeats" number input? Of course, it would not apply if you run out of entries, but still.

[ICodeDiffusion]

I agree it's really annoying, I don't want combinatorial feature but also don't want excessive repeats.

[eShep4]

I've been having this problem using a HUGE list of wildcards - the same ~20 of them keeps showing up over and over again.

I did a little digging, and the trouble lays in the bias in pseudorandom number generation in the random.choice() method. This is compounded by the pseudorandom seed being reset to the image seed each time - so the first option on the pseudorandom table is used, which tends to be one of a few options.

This seems to happen in _get_random() in the randomprompt.py library. Way too rusty on my coding to formulate a solution, but I think not-setting the seed and reshuffling the list each image should make it more subjectively random.

[CamiloMM]

That can't be right, can it? I assumed it was psychology messing with me. Even a crappy pseudo-random generator shouldn't have a bias that you would notice.