[Discussion] Factorio 2.0 support

[veger]

As Factorio 2.0 is released (quite soon from a software development point of view), I guess we need to prepare for this as well.

Some of the announced features will have (quite) some impact, I am thinking about Quality for example. For the other things I wonder if the impact is big (or if there is any).

Let's use this ticket to brainstorm a bit and figure out our plans with respect to Factorio 2.0 and maybe make a list of (breaking) changes with impact here?

[SWeini]

2.0

Expensive Mode Expensive mode is gone, so the code that reads RecipePrototype and TechnologyPrototype need to be adjusted.

Trigger Technologies Milestone analysis and similar code needs to read technology triggers from the prototype and respect that.

Beacons Beacons will define how they stack with each other. Need to read the prototype and apply the effects appropriately. It's still a bit unclear to me how this definition of stacking applies to different type of beacons.

Space Age

Quality

• Restrictions of quality modules, probably API adjustments in EffectTypeLimitation and Effect
• Being able to select buildings of all available qualities with correct stats.
• UI adjustments, probably don't want to list different qualities of same item as different items. Maybe do something similar as for the fluid temperatures.
• Render quality icon overlay on top of item/building icon.
• Recipes can be either "don't care about quality" or "quality exactly X". How to select this?
• How to calculate item cost of higher quality items? Only the item-ingredient part of recipe cost is affected by quality (not fluids, power, size, throughput, ...), so the ratio of cost regarding different quality can be different per item.
• Quality levels can be defined in prototype stage, all of the above needs be dynamic.

Asteroid Collector There is probably a new prototype. How similar will it be to MiningDrillPrototype? What data defines the distribution of different asteroids? For larger asteroids you spend ammo to break them down before you can grab them. How do we want to model that whole area?

Weight Rocket logistic will rely on items having weight. Should somehow be used in throughput stats. "e.g. 1.5/s = 1200s per stack = 0.1 yellow belts = 20000s per rocket"

Fusion Reactor The neighbor bonus of the fusion reactor will be a lot more complex than a simple "2x5" grid configuration. Also according to the FFF the effect of neighbor bonus is higher plasma temperature instead of more of that. Definitely different from a normal reactor. Maybe even a new prototype.

Spoils Probably it's too much to make calculations based of that, but since some science packs are more efficient when less spoiled, this effectively influences the whole cost analysis. cost analysis currently uses the total amount of required science packs as a basis, but with spoils you need more of some packs than in theory. Do we want to allow the user to define the "default spoilage" of certain items?

Stack Inserter The new stack inserter changes the items/s a belt can transport. The stack inserter itself will also need a different calculation of items/s. Not sure how much realism there is currently (inserter capacity 2 is 2x the speed from box to box, but not when belts are involved). How much accuracy do we want to have?

Planets How should we handle different planets? There might be buildings that can only be built on certain planets. Do we want the user to be able to assign a planet to each sheet? Is it enough to have the planet-specific science pack as a milestone, or is "being able to travel to planet X" also interesting as a separate milestone?

Additional Productivity CraftingMachinePrototype already has a base_productivity. I remember some statement from a FFF that there will be a (configurable) max productivity bonus due to recycling. And there will be per-item (or per-recipe?) bonus productivity as technology effect.

Recycling I assume the recycling machines are just furnaces in a trench coat with no extra work, but who knows?

[DaleStan]

You covered everything I'd thought about on my own, and several things I'd forgotten. Here are my thoughts:

2.0

Expensive Mode :+1:

Technology triggers Accessibility and milestone analyses already support using any FactorioObject as a source for items. There isn't a way to configure this externally, but I have local changes that make SE's arcosphere collector a source for SE's arcospheres, and the accessibility calculations work properly. Cost and automation analysis aren't as happy, though. If we extend this to technologies, we can say that Oil processing is unlocked by Mechanics.mining.basic-fluid-crude-oil, and that the rocket launch trigger depends on Mechanics.launch.item-name or both Special.launch and Item.item-name, depending on how it's implemented in-game.

Beacon transmission effectivity For base game beacons, this should be a new array in EntityBeacon and a relatively simple change to ModuleTemplate.GetModulesInfo. YAFC doesn't currently support crafters with multiple types of beacons. My best guess is that for multiple types, a crafter finds all the beacons that affect it and then does beacons.Sum(b => b.profile[beacons.Count] * b.transmissionEfficiency * b.modules.Sum(m => m.effects)).

Space age

Quality Quality is both super simple and super complicated:

• I read somewhere that the output quality is based on the lowest quality input. If that's true, we can just multiply the number of recipes, items, and crafters by 5 and most of the things we have to deal with are done. We just read one prototype from data.raw and generate five items, recipes, or entities.
  • Except this is actually a bad idea, at least from the usability POV.
• I think we care about quality in eight of the eighteen cases: Modules, Inserters (for inserter selection in the preferences), Mining drills (built-in productivity? How does this stack with prod modules and mining prod research?), Crafter speed and power, Solar panel and accumulators, reactor/boiler/steam engine speed or prod bonus, beacon transmission efficiency. This means Module and most derived types of Entity.
• My preference is to have the model mostly ignore item quality. If we do this:
• Screenshots from the Factorio team make it appear that the machines treat each quality output as a separate recipe, so we'll need to do the same to get accurate building counts.
• Do higher-quality fuels have more energy? It seems like a logical extension, but I haven't seen any mention of that.
• When changing a recipe's crafter, we don't want to display all 15 possible assembly machines (or up to 55 for Bobs), so we probably need two sections, one for the crafter and one for the quality. Something like this seems like a reasonable UI: As a stretch goal, we could display milestone icons for quality levels 2-5. However, all four levels unlock simultaneously, so I'm not sure how interesting that would be.
• The module selection dropdown will need a similar row. Most mods could use three sections, for type (speed, prod, etc.), rank (1-3 for vanilla, 1-9 for SE), and quality, but I think we'd need a lot of mod support for that. Many mods may support the third section with the module_category property, but some categories may only have one module (EI's Guass module, for example) and some mods may have too many categories to be feasible in the dropdown.
• Select*ObjectPanel will need an option for a similar row, depending on why things are being selected. It can't be automatic; NEIE and Dependency Explorer never care about quality.
• Quality modules change the recipe's output in ways that should be easy to handle when setting up the solver's inputs; it should work the same as (other) probabilistic recipes.

Asteroid Collector I hadn't thought about this at all, but my current thought is to treat it like a miner, that produces output from nothing. They have diminishing returns depending on count and variable returns depending on platform speed, but I don't have any good ideas for modelling that. Maybe just a setting and a notice like we currently have for fission reactors? It would be nice to model ammo consumption for breaking larger asteroids, but we don't have any similar models, e.g. energy/ammo consumption for killing biters, or repair pack consumption for the walls.

Weight :+1:

Fusion Reactor I think this is a separate setting (so fusion and fission can be configured separately) for 1xN or 2xN, along with radio buttons for "rectangular" (max 3 neighbors, as for fission), "offset" (max 4 neighbors) and "staggered" (half have 3 neighbors, half have 5 neighbors).

Spoils I'd overall prefer to ignore this, but one thing that we could do is designating one member of a link as "this member consumes 25% of the total output", to account for 25% spoilage. I haven't thought about how to get the solver to handle this.

Stack Inserter :+1: This probably "just" needs in-engine modelling, along with modelling the inserter improvements in FFF-419.

Planets I'd put this as low priority work, but it's also work that can be done and tested immediately. We can implement this for SE (for surface types) or Exotic Industries (for specific surfaces), and it should be pretty accurate to what Space Age will require. In both mods, some resource entities are only present on certain surfaces, and some crafting entities can only be built on certain surfaces. That said, I tried briefly in SE to lock Mechanics.mining.basic-solid.se-holmium-ore behind Technology.se-rocket-launch-pad, but the only interesting recipe that uses holmium ore is already locked behind a technology that depends on that technology.

Milestones have to be FactorioObjects, so "Access to this (type of) surface" would have to be a synthetic FactorioObject, probably with changes to the dependency analysis to allow another type of dependency.

Additional Productivity Configuring the productivity techs this will be a pain for the user, but we can just spit out a VirtualScrollList with one item for each productivity technology, and then alter Product.GetAmount() to be aware of those settings and use the right one. (per-item is easy, per-recipe means each Product needs to know what Recipe it belongs to.) YAFC already uses base_productivity, but it's probably worth checking to see whether it uses it correctly.

Recycling The recycler is probably a furnace in a trench coat, yes. This will require improved detection of special recipes, so we treat (most) recycling as an item source in the same way that we treat barrel emptying as a fluid source (see #134). The engine could reverse the single-output recipes, but then the recycler would have to figure out whether to decompose SE's Low Density Structure using the initial recipe (copper, steel, plastic, and glass) or the improved recipe (beryllium, steel, plastic, and glass.)

[veger]

You guys thought of much more changes than I initially had in my mind, very nice! (and very scary, as we need to change quite a bunch of things to make Yafc at least usable, let alone fully covering all the changes...)

I started to comment on your ideas already in my mind, but then realized this becomes messy fast... Should we split this off into separate tickets/discussions first. To stay on topic on each subject?

[SWeini]

I would keep it all in this one issue until we have more information about API changes and can start implementing and testing it. Quality definitely deserves its own issue later on, it's just too big. The other topics could be split off when we have more information about API changes, can play around with it in the real game, and the discussion becomes more focused on possible UI changes. I don't know how much sense those detailed discussions make before we know the facts.

Just for keeping track of possible topics that need adaption it's not worth to split this into a multitude of issues, in my opinion.

[DaleStan]

2.0

Module Categories Via a post by KiwiHawk in the Omni discord, we have news of allowed_module_categories:

allowed_module_categories will be a field on every machine that takes modules

This seems like it shouldn't be too hard to implement. It may also help with some of the pY module issues.