It's not possible to override the types with using interface merging. However you can override the types in a class. I've done that (my CreepRole.ticksToLive is number rather than number|undefined. I've only tried it with a Cyclic definitions but in theory it should work with dynamic imports. It's ugly for sure but you only have to do it once.

//room.extra.ts
export class RoomExtra extends Room {
  get allCreeps() {
    return this.find(FIND_CREEPS) as import('creep.extra').CreepExtra
  }
  get hostileCreeps(){
    return this.find(FIND_HOSTILE_CREEPS) as import('creep.extra').CreepExtra
  }
  get myCreeps() {
        return this.find(FIND_MY_CREEPS, {filter = (c) => !c.spawning}) as import('creep.role').CreepRole
  }
  get spawningCreeps() {
        return this.find(FIND_MY_CREEPS, {filter = (c) => c.spawning}) as import('creep.role').CreepSpawning
  }
}

// creep.extra.ts    
class CreepExtra extends Creep {
  room: import('room.extra').RoomExtra
  ttl: number
  spawning: false
}

// creep.role.ts
class CreepRole extends CreepExtra {
}

class CreepSpawning extends Creep {
  ttl: undefined
  spawning: true
}

There really should be a typescript-starter++ that sets up best practice for extending the base types. It could also include all the most common extensions like per structure accessors.

I only use the inheritance chain for Creep. It has completely eliminated all this casts from my code.

For all of my other modifications I do it with a single class in a single file.

I also have serveral mixins shared among different types. I keep the mixin class and interface declarations in a single file.

My Room extensions are kind of a mess half are in a single file the rest are sprinkled wherever and all of the declaration are in types.d.ts. I'm not happy with it but it hasn't climbed my annoyance list enough to fix it.

Here is my current solution.

I have several creep extension classes spread across different modules that I inject into Creep.prototype.

function mergePrototypes(klass: any, extra: any) {
        const descs = Object.getOwnPropertyDescriptors(extra.proto);
        delete descs.constructor;
        Object.defineProperties(klass.proto, descs);
}

function injecter(klass: any) {
    return function (extra: any) {
        merge(klass, extra);
    }
}

@injector(Creep)
class CreepExtra extends Creep {}

// or if you don't want to polyfill class decorators
class CreepMove extends CreepExtra {}
merge(Creep, CreepMove);

// ...

class CreepRepair extends CreepCarry {}
merge(Creep, CreepRepair);

For the most part the types just work. I have a few methods that I needed to explicitly interface merge in to prevent cyclic imports. I've debated whether to merge all the extension classes into a single file, but so far I haven't needed to. Occasionally I need to type cast this.room.find(FIND_MY_CREEPS) as CreepMove[], but I hasn't become enough of an issue.

The other big advantage of this mode is that VSCode go to definition takes me to the code rather than the interface.

Creeps can move past each other, but Creep.moveTo will not try to by default. The simplest solution is adding the ignoreCreeps option, but that has other other disadvantages. Traveler.js is a good example of improving on moveTo. You can use it as is, fork it as a starting point for your own movement code, or as inspiration for your own.

I don't play on shard3. Some of the assumptions in this question are foreign to me. The specialization and codependence is something that doesn't really show up on shards with more CPU. Sure we rely on each other to balance our mineral I/O, but by GCL 7 you have can build any tier3 boost you want independently. If the power market is too expensive I can sacrifice some rooms to give me enough CPU slack to run power miners.

I can't imagine getting to GPL 52 with only 2 rooms. How long would that take even with 100% power processing up time?

@x3mka I read the code. Only one can extract at a time. Dual harvesting works with minerals because the current cooldown is tracked as _cooldown and later it's applied to the actual .cooldown. Deposits update the cooldown directly.

update: multi-harvesting would still work before cooldown starts to occur. But if my math is correct that's around 500 extracted (so pretty early on).

@o4kapuk I think this should be a bug. Dual harvesting is an interesting optimization that some players make for mineral harvesting. It'd be a shame if they could not do it as well for Deposits.

I like the idea, but it will be a mod for private servers, more than a feature of the core engine. @ags131 and his screepsplus experience would be a good resource if this was something you'd like to pursue.

@communication The tone of your comment seems harsh and troll-y. Like when someone asks a question to vent their frustration rather than looking for a real answer.

But lets assume I'm misreading the tone and answer to parts.

What is the point? Philosophy for a moment, why add new features to a game at all? On a less deep plane I think the point is to enrich the breadth of screeps. Create so much content and possibilities that no one can write code for it all. In that messy, lovely sandbox players can choose which aspects that appeal to them most and long time active players still have new challenges to conquer.

What do factories change? How do the devs foresee factories interacting with existing systems, (also acknowledging that a good system should be solid and vague enough to allow new emergent, unforeseen play).

My impression is that factories will help anchor Strongholds into the world better.

The Deposits are geodiverse, which is a new element of the game (until now we've really only had center, meridian, edges, and corners for sector variation).

I believe elsewhere in forum @o4kapuk said that one of the goals was to encourage diversity in rooms for empires >10 rooms. Up to now there has been very little pressure to specialize rooms.

The Deposit spawn, harvesting, and decay are tuned to defeat any single obvious best approach.

An finally wrt to the economy which I think might be your real question. Why are NPC buy_orders being removed from the market? Why not keep them both? To that I only have speculation. On the surface limiting buy_orders to Factory commodities helps anchor Factories into the world better. There might be some other problem the devs a trying to correct with the market, thus killing two birds with one stone.

With all the recent news regarding ping back vs redirects the redirect might be uBlocks fault as well.

If you disable ping backs most sites use a redirect instead. All browsers except Brave are removing the ability to disable ping back because it only results is poor experience with no added privacy. It's trivial for sites to detect and switch to redirects when ping back is not working.

You really should allow ping backs in uBlock, your browsing will be faster.

Now that the tracking is less obtrusive you can go on to wonder why checkload.dir is tracking you.

screen shot of google using a ping back. With uBlock disabling ping back the url will be a redirect instead of a direct link.

@GimmeCookies I'm sure there is a lot of sarcasm there, but lets be clear.

I'm paying for Screeps MMO. Screeps Inc (or whatever) gets my respect for open sourcing the private server with a permissive licence and keeping it up to date. Part of the value proposition is that they also maintain the opensource, however I am not paying for the opensource repos. Further to the point, the more we as a community can drive the opensource the more it becomes defacto ours.

If Screeps Inc ever becomes non-viable (like if Artem boards a rocket headed to Mars), the repos will remain. If we ever loose access to screeps/* repos, I vote ags131's forks become the preferred upstream. At that point I'd encourage everyone to join @ags131 's patreon. Then I would be paying for the opensource.

In case it wasn't clear, I am really happy with Screeps Inc's handling of the opensource. I don't need anything more. I wish we had more, but I also wish that Screeps was as huge as EveOnline and a core part of every highschool's compsci curriculum, and that Artem turned into some fat CEO behind a desk managing a hundred person team wondering if he'll ever have time to get back to the code.