I'm really sorry about this

[davidn]

Apparently not feeling like he had unleashed enough suffering on the world, Qrostar - the creator of Hanano Puzzle - has released another puzzle game. This one is called Jelly Puzzle, or something like it, and will be instantly familiar to anyone who played the previous one - it features more coloured blocks and infuriating PxTone music in prime-numbered time signatures, this time with a higher-resolution Worms 2-ish aesthetic.


HOW THE HELL ARE YOU MEANT TO DO THIS

Have fun with this one, because this time around, non-Japanese speakers don't even get the benefit of instructions! Left-click to move a block left, right-click to move it right. Blocks of the same colour (blue, not-blue and other) placed adjacent to each other merge together. That's all I've been able to work out so far. The solution to level 1 is not among these things.

I'm not 100% sure as to the goal of the game yet (though it's probably to eliminate all the 1-block pieces), but after the last one, I think the only thing this author would ask you to do with jelly is attempt to nail it to the ceiling.

Get it here or here - and say goodbye to your brain.

Comments

[premchaia_pre4]

I decided for tradition's sake to mechanical-search level 19, in C instead of Lua this time—and as before, it found that I had gotten the inferred constraints of the solution exactly right but had missed one sequence that actually made them possible. Level 20 was easy. There.

Whrf!

[davidn]

Oh, so by brute-forced you didn't mean you just manually tried every available move...! What kind of thing are you writing to solve them?

[premchaia_pre4]

The kind of opaque C that I would avoid knowingly putting into any kind of real production application!

Er, which is to say something like this. Or there's the Lua for Hanano Puzzle ^#29. I'm told that academic scientific programmers tend to write like this even when they're not half-asleep and doing complete one-offs, which is a bit scary…

(I did both of these, as I recall, without having had any coffee yet that day, respectively, which tends to leave my thoughts kind of meandering.)

[davidn]

Do you... realize that you re-wrote the entire mechanics of these games (minus interaction and display) in one morning each, pre-coffee? That's... not a normal thing!

I'm impressed with the jelly19 engine... I think I would only have been able to do it in a much more verbose way. Your C code is actually much more understandable than most I've seen, though, mostly due to the presence of vowels.

[premchaia_pre4]

Well, they're not particularly complicated games in terms of actual physics!

hana29 is actually a strict subset, since I didn't need more than one wide block, more than one flower, or more complicated shapes for just the one part of the one level. I think jelly19 implements the whole thing though, because level 19 of that game is complicated enough to need all the physics. (Well, it doesn't need multiple colors, but it felt better to put that in anyway.) I remember I ran it on level 20 afterwards (just altering LEVEL_DESC) out of curiosity, and it found a solution similar to the one that I'd used, but better-optimized in moves, as one would expect from a mechanical search.

Some of the opaque and/or sharp edges about the C are: my affinity for bit-twiddling showing through in the use of bitmasks for sets of small integers (the use of which is not described nearby and the limits of which are not enforced); the State type packing a block ID and a color into an 8-bit value resulting in some fragile shuffling of unsigned integers with unobvious interpretations; some crufty subexpressions that aren't factored out (like '1' + (foo - MIN_BLOCK_ID)); the weird superfluous union of a 2D and a 1D array in Board; and just a general lack of semantic documentation. But it's sort of to be expected in context, I suppose.

[davidn]

If you understand it and you don't expect anyone else to want to ;)

My own solver application (which I can now just use on arbitrary classes that extend State, which describe a game situation, the available moves, how to transition from one to another and whether this state is solved) started off in a rather similar way... it was written for a very specific puzzle (the Knight's Tour, actually, which I didn't look up the algorithm for). The class for sliding puzzles has now expanded to being able to handle puzzles with any shape and number of pieces, with any number of goal pieces.

[davidn]

(Which still counts as having solved it with your own ingenuity! So I tell myself every time I do the same for a sliding block puzzle in Professor Layton.)

[(Anonymous)]

*still JD*

If I have to suffer through the sliding puzzles in Professor Layton games then everyone else should as well!

Also I'm posting to say that I'm hopelessly stuck on puzzle 20 and then I'll just wait for the magic to happen...