Yesterday, I discovered regexle [1] and today found out that this was the inspiration for it.
I think I've seen this puzzle before, but I really enjoyed myself gradually solving it while doing other things in the last two hours. I'm sharing this for anyone else who wants a way to spend their afternoon!
Because it's so trivially unsolvable, I had a quick look at the posted solution.
We're reading it wrong.
Take the first regex on the top row, one you included in your example:
[^X]*(DN|TE|NI)
You're supposed to fill that in from the bottom to the top.
In other words, the first letter goes immediately to the right of [^c]*[^R]*III.* - that square has to be a 'not X'. The hex immediately below it has to be an E, I, or N.
That was pretty hard! It took me a bit over 2 hours. It took me a long time to get started, it felt impossible.
But I found a lot of satisfying "ah ah!" moments that kept me going.
Really well done, but now I don't ever want to do one of these again.
The worst is when you're an hour in and there's some crucial wrong assumption you've made at some point that's cascaded across the board and you have to start again. Had to come back to it a day later, but very satisfying to complete. I must be a masochist because I want to do another.
I'm down to just one unmatched expression, but it's going to require undoing many others to fix.
With so many wildcards (`.*`) involved, it feels as though making a few assumptions along the way is unavoidable. If I start over and only make moves I can prove are correct, I don't get very far.
This is a really cool idea! It seems really hard though, I tried for a while and didn't get far. Would be keen to see one that's a little easier, maybe with less * qualifiers and more + or ?
I tried this for over 20 minutes and made almost zero progress because I thought the expressions also had to be words, like a crossword puzzle. Oops. It might be worth clarifying that somewhere.
Yesterday, I discovered regexle [1] and today found out that this was the inspiration for it.
I think I've seen this puzzle before, but I really enjoyed myself gradually solving it while doing other things in the last two hours. I'm sharing this for anyone else who wants a way to spend their afternoon!
[1]: https://regexle.com
I like that one better. Smaller. More manageable. I don't want to spend 2 hours deciphering the big one.
Took 4.5 min to solve #519. Would have been quicker if I started with a better strategy but it was my first time.
I got #520 in 3m 8s with some practice on the big one.
I thought the bigger one was more fun though, I'll continue with that one now.
Got 3m 32s on #519!
#520 seems much harder tho!
What a great game, just wish the colours were configurable (colour blind)!
519 seems to have at least 3 solutions... feels a little underconstrained!
It seems impossible?
On the left a horizontal must start with F: F.[AO].[AO].*
However, the diagonal for that cell may not include an F (aside from the start): [^X]*(DN|TE|NI)
BIG EDIT:
Because it's so trivially unsolvable, I had a quick look at the posted solution.
We're reading it wrong.
Take the first regex on the top row, one you included in your example:
[^X]*(DN|TE|NI)
You're supposed to fill that in from the bottom to the top.
In other words, the first letter goes immediately to the right of [^c]*[^R]*III.* - that square has to be a 'not X'. The hex immediately below it has to be an E, I, or N.
[^X] means any character other than X, so [^X]* matches zero or more of any characters, so long as none of them are X.
Ah my bad! I did misinterpret [^X] as X or the line-start.
[^X]* means anything but X as many times as you want. So F will match this part.
That was pretty hard! It took me a bit over 2 hours. It took me a long time to get started, it felt impossible. But I found a lot of satisfying "ah ah!" moments that kept me going.
Really well done, but now I don't ever want to do one of these again.
The worst is when you're an hour in and there's some crucial wrong assumption you've made at some point that's cascaded across the board and you have to start again. Had to come back to it a day later, but very satisfying to complete. I must be a masochist because I want to do another.
I'm down to just one unmatched expression, but it's going to require undoing many others to fix.
With so many wildcards (`.*`) involved, it feels as though making a few assumptions along the way is unavoidable. If I start over and only make moves I can prove are correct, I don't get very far.
Edit: Solved! :)
> Edit: Solved! :)
:)
https://blog.nelhage.com/post/regex-crosswords-z3/ was posted a few days ago here and is an interesting way to solve these
Yeah, that's how I found regexle, which eventually led me to this!
This is a really cool idea! It seems really hard though, I tried for a while and didn't get far. Would be keen to see one that's a little easier, maybe with less * qualifiers and more + or ?
I tried this for over 20 minutes and made almost zero progress because I thought the expressions also had to be words, like a crossword puzzle. Oops. It might be worth clarifying that somewhere.
For something a bit simpler I made redoku a while ago: https://padolsey.github.io/redoku/
Great puzzles! Didn't have so much fun puzzling for a long time.