Symbols gets sorted even for user types

[JanisErdmanis]

I want to implement some noncommutative algebra rules for revising lengthy and thus error-prone calculations, as explained in the discourse. Currently, I am facing an issue that the symbols, even for user types, are being sorted automatically by their names. Which, for instance, prevents me from applying a rule:

abstract type SU2 end
@syms ηx::SU2 ηy::SU2 ηz::SU2
(@rule ηy*ηx => -im*ηz)(ηy*ηx)==nothing

[shashi]

Yes this is because there’s an issue which required us to define * on Symbolic rather thsn on Symbolic{Number}

You can define the method on Symbolic{SU2} to construct a Term instead of a Mul which would avoid the sort.

Should get closed with #155

[JanisErdmanis]

Ok. So the rule indeed works on

(@rule  ηy*ηx => -im*ηz)(Term(*, [ηy, ηx])

However, if I try to automate the process by defining a method for * on Symbolic{SU2} like:

*(x::Symbolic{SU2}, y::Symbolic{SU2}) = Term(*, [x, y])

the rule matching fails with an error message:

Failed to apply rule ηy * ηx => -im * ηz on expression ηy * ηx
  (::SymbolicUtils.Rule{SymbolicUtils.Term{Any}, SymbolicUtils.var"#term_matcher#76"{Tuple{SymbolicUtils.var"#literal_matcher#72"{typeof(Main.workspace40.:*)}, SymbolicUtils.var"#literal_matcher#72"{SymbolicUtils.Sym{Main.workspace3.SU2}}, SymbolicUtils.var"#literal_matcher#72"{SymbolicUtils.Sym{Main.workspace3.SU2}}}}, Main.workspace40.var"#51#52"})(::SymbolicUtils.Term{Any})@rule.jl:142
    top-level scope

Are there a better way?

[shashi]

Okay now we throw the appropriate method errors and the usual operations are only defined on Symbolic numbers.