first draft of crafting combinator article

submissions/crafting_combinator/crafting_combinator.md

@@ -0,0 +1,276 @@
+## Crafting Combinators: Filling a hole in the circuit network <author>Max Baroi</author>
+
+The circuit network is a strange and rich subsystem. 
+It's relatively early in the research tree, but is more of a hallmark of 
+later and more advanced bases. 
+Many players completely eschew the topic; only using a lone decider combinator 
+for the most basic of fluid balancing. 
+Others hear there's a Turing complete mechanic and take that as a challenge.
+For readers unfamiliar with the concept, in practice Turing completeness
+means one can and *will* [implement Doom](https://www.youtube.com/watch?v=28UzqVz1r24).
+Though we haven't seen any attempts to port 
+Doom to [belts](https://alt-f4.blog/ALTF4-23/).
+
+
+The [list of entities](https://wiki.factorio.com/Circuit_network#Devices) 
+which can be connected to a circuit network is comprehensive. 
+A fair number of readers have probably hooked up an inserter to prevent a chest from overfilling, 
+or set a train stop limit. 
+But how many have *set* rail signals to squeeze as much throughput as possible from 
+an unbuffered four-way intersection? 
+The answer is more than zero and a trip to 
+[Railway Setups](https://forums.factorio.com/viewforum.php?f=194) 
+or [Train Help](https://discord.com/channels/139677590393716737/760455214049263646)
+will expose you to arcana you didn't know existed. 
+But notice what's conspicuously absent from the above list: assemblers, chemical plants, 
+refineries, and centrifuges. 
+Players can only influence these producers indirectly by manipulating what feed or empty these 
+producers. 
+Luckily, theRustyKnife's 
+[Crafting Combinator](https://mods.factorio.com/mod/crafting_combinator) mod plugs in this gap. 
+It introduces an eponymous combinator that can either read a producer's recipe or 
+set it, along with a recipe combinator which conveniently provides the ingredients or recipes
+associated with a given item. 
+
+Our goal is to write the section of the 
+[circuit network cookbook](https://wiki.factorio.com/Tutorial:Circuit_network_cookbook)
+which would need to be added if crafting combinators were in vanilla. 
+And then maybe one or two more examples which we think are cool. 
+Of course, one can delve deeper and start to apply more advanced concepts 
+(like those found in the wiki's 
+[combinator tutorial](https://wiki.factorio.com/Tutorial:Combinator_tutorial)
+or pocarski's previous 
+[posts](https://alt-f4.blog/ALTF4-37/)) in conjunction with crafting combinators.
+Players like gridstop have taken the idea to the extreme conclusion and launched rockets
+from a 
+[single-assembler factory](https://forums.factorio.com/viewtopic.php?f=193&t=42964). 
+But for my purposes, we will avoid anything more complicated than an RS-latch.
+Even then, I'm sure anyone on the 
+[combinators channel](https://discord.com/channels/139677590393716737/783049079584063588)
+or [combinator creations](https://forums.factorio.com/viewforum.php?f=193)
+will explain why my circuit logic is bad, and I should feel bad.  
+
+# A First Example
+
+First consider a set of items which have an identical ingredient list; mining drills and basic assemblers. 
+These are relatively expensive in the early game, 
+but are not entities which are continuously consumed.
+Let's make a single factory which makes sure a chest contains 
+five mining drills and five basic assemblers.
+
+![Miner/Assembler setup](basic.gif "Basic crafting combinator setup")
+
+Let's walk through the logic with the following diagram as an aid.
+
+![Logic Breakdown](basicdiagram.jpg "Basic Crafter Diagram")
+
+The constant combinator sets the demand for the items. 
+We have a wire leading from the chest storing our outputs which negates the quantities. 
+The idea is the result of subtracting *current supply* from *desired supply* is precisely what we have
+left to manufacture. 
+The new bit is the crafting combinator itself. 
+It is in write mode, and under misc is the default options which we explore (and modify) later.
+Given two competing signals, a default crafting combinator in write-mode will set the
+producer recipe with the largest signal strength. 
+The chest fills, and the signals decrement until they (ideally) reach zero when the set demand is met   
+and the producer shuts down. 
+We have the decider combinator to only pass positive-valued signals to the crafting combinator. 
+Even in our basic example, the production run ended with six miners even though we only asked for five.
+With stack inserters, fast crafting speed, and signal propagation delay we will usually over-produce by
+a few items. 
+In this case, the circuit will contain a miner signal with value -1. 
+Negative-valued signals are still signals, and activate the crafting combinator. 
+Without the decider combinator, the assembler would not stop, and we'd end up with a chest with five 
+basic assemblers and 2,350 miners. 
+Five is closer to six than 2,350, hence the decider combinator. 
+Filtering for positive-valued signals will be an extremely common idiom in 
+our crafting combinator setups. 
+
+# The Recipe Combinator
+
+Of course, the circuit network components themselves form a set of items with identical 
+ingredient lists. 
+![Circuit Line](circuitline.jpg "Circuit Network Components Assembly Line")
+Asides from speakers, the other ten items (we can't forget the crafting and recipe combinators) 
+are all formed from electronic circuits, copper cables and iron plates. 
+This feels like a perfect use-case for crafting combinators. 
+But instead of just returning to our original example with a different 
+demand list, we will add a complication. 
+Transporting iron sticks and copper cables is a sin against throughput. 
+We should directly insert these intermediaries, and only supply iron plates, 
+copper plates, and electronic circuits.
+We make no claims the following complex is the best way to accomplish this task, 
+but we do construct a complex with two assemblers which 
+can craft eleven different objects with direct insertion of low density 
+intermediaries. 
+And we did so without implementing pushdown automata.
+All kudos to nano factory architects, but I can only cognitively handle a buffer chest or two. 
+
+![Circuit Network Components](circuits.gif "A Crafting Combinator Crafting Combinators")
+
+We only included constant combinators and speakers because the interesting part of the 
+setup is when we switch to items with slightly different ingredients. 
+So let us break down what is happening. 
+
+![Circuit Crafter Breakdown](circuitsdiagram.jpg "Circuit Crafter Diagram")
+
+The little circuit setup on the far left with the constant combinator is the
+logic used in the previous example which measures current supply against 
+desired supply to determine what item the assembler should craft. 
+What is now different stems is the crafting combinator at position `B`.
+This crafting combinator is set to *read-mode*, and passes along to the 
+circuit network what the assembler is currently trying to manufacture. 
+This is then fed to our new recipe combinator at `C`, which outputs 
+the recipe for the input.
+So when the assembler on the left is producing constant 
+combinators, the output to this recipe combinator is five copper cables and two electronic.
+When the assembler is producing speakers, the output is five copper cables, four 
+electronic circuits, three iron plates, and four iron sticks.
+We only want the ancillary assembler to manufacture either copper cables or 
+iron sticks. 
+The arithmetic combinators at `D` and `E` are there to filter the recipe signals
+and only pass through the signals for intermediary products which we want the 
+ancillary assembler to produce. 
+We add an arithmetic combinator at `A` which negates the count of the intermediary
+products which reside in the buffer chest.
+Then the combined values of the signals from `A`, `D` and `E` are how many copper cables 
+and iron sticks the ancillary assembler needs to produce. 
+The decider combinator at `F` only allows positive signals for copper cables and iron sticks 
+to be passed to our crafting combinator at `G`, which is set to write-mode. 
+
+# Production Priorities 
+If we want a single assembler to produce ten units of `A` and five units of `B` there 
+are a few natural orders which we could desire.
+I apologize if there's some industrial engineering terms which are standard but of
+which I am ignorant.  
+
+| Priority | Production Order |
+| - | - |
+| The Largest Deficit | AAAAAABABABABAB |
+| One at a Time | AAAAAAAAAABBBBB |
+| Interlaced | AABAABAABAABAAB |
+
+So far we have only done the largest-deficit method. 
+We measure the difference between 
+current supply and desired supply and signal corresponding to the item with 
+the highest value is passed to our assemblers. 
+In this instance we produce five units `A`, at which point we alternate between the
+`A` and `B`. An alternative is to produce ten units of `A`, and only once `A` is satisfied, 
+we start producing `B`.
+We can also interlace production, so we produce two units of `A`, then one unit `B`. 
+This way the current supply of items is roughly proportional to the desired supply of 
+items. 
+I produce way less oil refineries than miners, but I might not want to wait to 
+produce hundreds of miners before my factory starts restocking oil refineries. 
+
+As we have seen, crafting combinators default to the largest-deficit strategy.
+The "keep crafting until zero" option in the combinator interface makes 
+crafting combinators adopt the one-at-a-time strategy. 
+But in my experience, I usually want an interlacing strategy.
+To focus on our new problem let's return to a set of items which identical 
+ingredients: logistics chests. 
+Suppose we want logistic chests in stock at the following quantities:
+
+| Chest | Quantity |
+| - | - |
+| Passive Provider | 100 |
+| Requester | 100 |
+| Storage | 50 |
+| Buffer | 20 |
+| Active Provider | 10 |
+
+
+
+![Chest Complex](https://youtu.be/82_pvGBz2BE "Logistic Chests Assembler")
+
+We produce the chests at the desired quantities using an interlaced 
+production strategy. 
+Let's trace the logic behind this setup.
+
+
+![Chest Crafter Breakdown](chestdiagram.jpg "Circuit Crafter Diagram")
+
+The series of combinators on top attached to the filter combinator is uninteresting. 
+We include it to transfer only finished products from the overflow chest 
+to our final storage chest. 
+As usual, the constant combinator determines the desired supply.
+The highlighted arithemtic combinator at point `A` calculates 
+the ratios of all the desired items to the item in the highest demand. 
+The idea is for every passive provider chest, we want to produce
+one requester chest, two storage chests, five buffer chests and 
+ten active provider chests. 
+The output of the arithmetic combinator reflects this desire. 
+The chest at `B` is our current supply of items. 
+We multiply the outputs at `A` and `B` on a per-signal basis by adapting the
+[multiplier](https://wiki.factorio.com/Tutorial:Combinator_tutorial#Multiplier_and_Dictionaries.2FArrays)
+from the wiki. 
+This accurately counts passive provider and requester chests, 
+but double counts storage chests, quintuple counts buffer chests, and
+dectuple(?) counts active provider chests. 
+We subtract this quantity from 100 (the highest quantity demand of any chest),
+and the highest resulting signal is what is passed to the
+crafting combinator. 
+Of course, we filter for positive signals as we always do. 
+
+
+If this is a little abstract let's work through a specific example
+with only two chest types. 
+Suppose we have fifty passive provider chests, and twenty storage chests already. 
+We have 50% of our desired passive provider chests but only 40% of our desired storage chests,
+so we want to prioritize storage chests.
+In this example, my `100-AB` quantity for requester chests is 100-50*1=50
+and for storage chests is 100-20*2=60, so the crafting combinator will set the
+assembler recipe to storage chests as desired.
+When will these two `100-AB` quantities be equal?
+They will be equal once we have twenty-five storage chests. 
+At that point we will have 50% of our desired passive provider *and* storage chests. 
+Production dynamically adjusts to maintain our desired item ratios. 
+
+# Putting the Pieces Together
+I don't want anyone to think that crafting combinators are only for 
+the early or mid-game; they're just a method to save some space. 
+So let's build complex worthy of a later-game mall using the all 
+the concepts we've covered so far. 
+
+We build a mini-mall stocked with the following items at the following amounts
+| Item | Quantity |
+| - | - |
+| Mining Drills | 1000 |
+| Assembling Machine 2 | 100 |
+| Chemical Plant | 200 |
+| Oil Refinery | 100 |
+| Offshore Pump | 100 |
+| Pumpjack | 100 |
+
+Of course, in a real late-game mall, we wouldn't store assembling machine 2s but use them
+as ingredients for assembly machine 3s. 
+There will be a side complex to produce intermediary products on demand 
+using a recipe combinator, and we will use our multiplier technique
+to have an interlaced production strategy. 
+Passive provider chests will be used as overflow chests, so we can recycle 
+materials that would otherwise be discarded. 
+Final products will be stored in single-item buffer chests to 
+prevent logistic network loops or storing intermediary products.  
+And look at what we have wrought together in so short of time
+
+![Mall Complex](https://youtu.be/wZR4nz1XpMg "Mini Mall")
+
+# The Road From Here
+Crafting combinators add a new mechanism to create a dynamic factory that 
+responds in real time to current supply and demand. 
+This article has only scratched the surface of what is possible mainly because 
+my ability is surface-deep. 
+There are plenty of designs I would like to explore: 
+
+- modules complexes which rebalance production given factory consumption
+- recipes setting recipes setting recipes
+- city blocks where production is set by trains
+- assembler recipes dictating belt pathways
+
+The main reason I wish Wube would bestow the 
+[ultimate honor](https://alt-f4.blog/ALTF4-21/) to theRustyKnife
+is I want to see what the community can do with crafting combinators. 
+There's a lot of talent out there, this blog highlights it on a 
+regular basis, and I want to be amazed at what the community would accomplish 
+if crafting combinator were added to vanilla. 
+