Boulders

[paulkent-um]

I'm going to be ignoring boulders for a while yet, because those are going to be very difficult to manage competently, and I have some improvements to the agent's behavior that have a better boost-per-man-hour ratio. But, I'm going to mark this as an open issue for the time being, in case you have any thoughts.

Facets:

• Boulders can be pushed to open up new areas. Sometimes this is faster than groping around the walls looking for a secret door. More rarely, there just isn't a secret door that can take you to the other side of the boulder.
• A boulder with a wall next to it can't be pushed perpendicularly to said wall, only parallel. If you push rocks carelessly, you can easily do more harm than good by blocking a vital path or intersection in such a way that the rock can't be moved back. Therefore, rocks need to be pushed strategically, and I don't know what such a strategy would look like. I would imagine there's an open-source Sokoban solver out there that we can steal gratefully inherit; but even then, most Sokoban games are fully-observed by the player, which this game isn't.
• Speaking of Sokoban, there is a dedicated Sokoban section of the game relatively early on, where pushing boulders is required and most possible ways to cheese the puzzles are punished with luck penalties. But the whole section is optional (the only things you stand to miss by skipping it are a few nifty items and a bunch of food), and even if it wasn't, the levels there aren't procedurally generated, so we could just look up the suite of possible layouts, write out solves for them, and basically hard-code said solves.
• A monster can block the path of a boulder, preventing it from going anywhere. This is distinguished from the push-against-wall case by a different message. Most likely, the only recourse in this situation is simply to try again later, which is a difficult idea to convey in my current architecture. I could add a third layer of map to the main map and the "areas searched" map, but is it really worth it?
• Alternatively, you can move past a boulder, but this requires dropping just about everything in your inventory, including all your weapons, protective gear, food, and useful tools. Even leaving aside the risk of a monster making off with the undefended goods (I don't think Elbereth can safeguard unattended items anymore?), we'd need a whole protocol to determine at what point it's better just to run back for your items.

[Rajathbharadwaj]

If we have the tools to break it into stones, it's better to do it. However, if we posses some unwanted things, it's better to squeeze past the boulders. Check this out, although I feel with the risks vs rewards, it's better to come back later. (https://nethack.fandom.com/wiki/Boulder#:~:text=Monsters%20cannot%20attack%20through%20boulders,spells%20to%20attack%20through%20them.&text=To%20get%20past%20a%20boulder,of%20striking%20or%20force%20bolt)

[paulkent-um]

The fact that we can attack over boulders does help, as does the potential to break it apart. There's also the option of Stone to Flesh, since Healers start out knowing it. So I suppose even without knowing a reliable way to figure out how best to push boulders, we still have options.

[Rajathbharadwaj]

The fact that we can attack over boulders does help, as does the potential to break it apart. There's also the option of Stone to Flesh, since Healers start out knowing it. So I suppose even without knowing a reliable way to figure out how best to push boulders, we still have options.

Yes. Btw what are your options? Stone-to-Flesh? That won't work all the time. :D

[paulkent-um]

Well, stone-to-flesh, break it apart into rocks, use a ranged weapon over it to clear out monsters blocking the way behind the boulder. That's three options right there, not even counting more unusual techniques like polyselfing into a giant and lifting the boulder.