gereleth

Joined 4 years ago

Karma

338

Submissions

Member since

Jul 24

View on Hacker News

Recent submissions

Comment · 3 years ago

This is a good start! Exactly like hexapipes began lol. I just wanted to make a solver to overcome my addiction to this puzzle.

Come open a discussion on github if you have further questions or just want to share your progress =) https://github.com/gereleth/hexapipes/discussions

Comment · 3 years ago

I think the downloaded version of Penrose puzzles does not currently contain enough info to run a solver on it. Because the grid is different every time and this info is not present in the downloaded file.

But I can tell you how it works for other grids =).

The contents of `tiles` key are an array of tile shapes. Possible directions where there are connections are binary-encoded. For example, on a square grid 1 means right, 2 means up, 4 means left and 8 means down. So a 3 means a corner piece connecting up and right, a 5 is a horizontal straight piece. A 0 means an empty tile, 15 in this case is a fully connected + tile. It's similar for other grids.

How the tiles are numbered depends on grid implementation. In a square 5x5 puzzle the indices would go like this:

0 1 2 3 4

5 6 7 8 9

...

You can look at grid implementations here: https://github.com/gereleth/hexapipes/tree/main/src/lib/puzz...

You can also import a downloaded puzzle (non-penrose) on the custom puzzle page and watch my solver work on it)). https://hexapipes.vercel.app/custom

Comment · 3 years ago

Hexapipes author here =). This is awesome! People kept mentioning Penrose grid to me but I didn't know where to even start with it. And then you went and made it happen! I admire the dedication it took, love collaborating with you!

The bumpy history of this grid's development can be found in this issue: https://github.com/gereleth/hexapipes/issues/83

First hexapipes thread for context: https://news.ycombinator.com/item?id=32216436

Comment · 4 years ago

Haha, way to go, this is awesome. Did it solve it all without human input? Can it do wraps?

Sorry about your sleep schedule).

Comment · 4 years ago

I see, thanks for reporting.

This contextmenu issue is sure showing me the ugly side of frontend development :/

Comment · 4 years ago

Congrats! This is some amazing dedication =).

Comment · 4 years ago

Sorry about this. I sometimes place them accidentally too.

I have an idea of placing them with some kind of gesture - like touch and drag between tiles. That wouldn't interfere with tapping to rotate tiles. But on the other hand, I'm usually zoomed in on mobile and I use touch and drag to move around. Interfering with that would be bad too.

I'm open to ideas for better controls.

Comment · 4 years ago

Thanks for the feedback. I'm very sure all puzzles on the site are solvable. But a skip button makes total sense when one hits a deadend. I'll work it in when I have the time.

With a randomize button do you mean a complete reset to initial state? Or something like keep the locked tiles, reset the rest?

Comment · 4 years ago

Browser zoom kinda works. But yeah, adding better zoom/pan is in my backlog.

You sure are brave to take on a 40x40 =)

Comment · 4 years ago

They all need to be connected into a single network. No hanging connections, no loops, no disjoint parts.

Comment · 4 years ago

Thanks for reporting, I'll look into this.

Comment · 4 years ago

Are you on a Mac? I thought I got rid of this but I haven't tested on those.

Comment · 4 years ago

On the backlog it goes). Can't promise I'll get to it in a timely manner but thanks for the suggestion.

Comment · 4 years ago

Oops, thanks for reporting. I'll try to find someone with a Mac to debug this.

Comment · 4 years ago

I didn't use any special coordinates, just some trigonometry. My code transforms coordinates between (x,y) - for drawing, (row, column) - for finding neighbours and (index into an array of tiles) - for managing state. It sort of grew organically and could probably be cleaned up a lot.

The hardest part was making the board repeat for the wrap variant making sure that all tiles have consistent wrap neighbours. I almost deployed a version that worked fine on odd-sized grids and went completely bananas on even-sized grids. The danger of always testing on a 5x5 board)).