MultisetSubstitutionSystem

[maxitg]

Changes

• Implements MultisetSubstitutionSystem.
• It supports arbitrary rules (e.g., conditions, a variable number of inputs on the left-hand side, and any expression on the right-hand side). Basically, it treats rules as black boxes.
• "MaxDestroyerEvents", "MaxGeneration", "MaxEventInputs" and "MaxEventOutputs" are supported for event selection.
• Only "MaxEvents" is supported as a stopping condition.

Comments

• It is currently very slow because it keeps trying to match the same sets of expressions over and over again. It will need to be optimized.
• Token deduplication is not implemented.
• Only one event ordering is implemented.
• It will be easy to allow using multihistory objects as inputs, however, it's not currently implemented.
• There is no rule parsing happening at all. This means the system does not even know how many tokens a rule input has (it just enumerates all subsets).
• @taliesinb, I'm now using "spacelike" -> "compatible", "timelike" -> "past" or "future", "branchlike" -> "inconsistent". So, the separations are now ordered relations. Let me know if there are any objections.
• The engine is procedural. It can probably be implemented in C++ for efficiency provided that we can callback Wolfram Kernel for matching and event instantiation.
• I did not implement self-consistency tests yet because they require time constraint and hypergraph system.

Examples

• Create a multihistory:

In[] := multihistory = GenerateMultihistory[
  MultisetSubstitutionSystem[{a_, b_} :> {a + b}],
  <|"MaxDestroyerEvents" -> 2, "MaxEventInputs" -> 2|>,
  None,
  {"InputCount", "SortedInputExpressions", "UnsortedInputExpressions", "RuleIndex"},
  <||>] @ {3, 8, 8, 8, 2, 10, 0, 9, 7}

• Convert to a WolframModelEvolutionObject:

In[] := wolframModelEvolution = SetReplaceTypeConvert[{WolframModelEvolutionObject, 2}] @ multihistory

• Produce its token-event graph:

In[] := wolframModelEvolution["ExpressionsEventsGraph", VertexLabels -> Automatic]

---

This change is 

[maxitg]

I did not quite finish the tests yet. I have half of the separation matching file remaining. Also, there is no documentation yet. I will make another PR with documentation for both GenerateMultihistory and MultisetSubstitutionSystem. Otherwise, this PR is going to exceed 1000 lines.

[maxitg]

@taliesinb, could you give me an estimate of when will you need this PR for the sequence system so that I know what my deadline is to finish this?

[maxitg]

@taliesinb, tests are done. The only thing remaining is figuring out the naming for spacelike/timelike/branchlike. I think the names I came up with, compatible/past/future/inconsistent are too confusing.

[maxitg]

Well, tests are almost done. Just need to fix a few things.

[maxitg]

I have finished everything here except for documentation (which will be in a separate PR together with GenerateMultihistory). Also, I went back to using spacelike/timelike/branchlike for now until/unless we come up with a better name. Since these are not public, they will be easy to change at any point.