Add Domains for variables. about modelingtoolkit.jl HOT 7 CLOSED

I would love to see this as the same one used for the differential operators DSL (which I think was your plan)

Yup, that's the plan. We should have a bunch of domain types and an untyped domain field for this. Our variable constructor probably needs to start using some kwargs.

from modelingtoolkit.jl.

In order of priority after continuous (unbounded) variables: I do not need any of the integer domains yet, and don't think we need compact subsets of R yet. But having a discrete domain would be nice within a few months (i.e. I have a system of differential equations with a discrete state i = :L or i = :H in the math where i have a Markov chain between L and H and then another continuous variable x. So the notation in my math is v_i(x), etc.)

from modelingtoolkit.jl.

I don't really understand enums well in julia, so it could be that it is better to use a finite domain of symbols, Fruits = (:apple1, :apple2, :apple3) and FiniteDomain(Fruits)

That's pretty much what an enum is BTW.

from modelingtoolkit.jl.

Perfect. Then I am happy with enums.

In fact, they may have the advantage that they have the default ordinal position attached to the names? If so it makes ordering the states of markov chains consistent without any other magic. i.e. @enum InnovationType H=1, L=2 then I can easily jump between markov chain symbols and the ordinal position within vectors.

from modelingtoolkit.jl.

In fact, they may have the advantage that they have the default ordinal position attached to the names? If so it makes ordering the states of markov chains consistent without any other magic. i.e. @enum InnovationType H=1, L=2 then I can easily jump between markov chain symbols and the ordinal position within vectors.

Indeed.

from modelingtoolkit.jl.

Domains have been added to the variables, but no domains have been created.

from modelingtoolkit.jl.

Basic domains have been added.

from modelingtoolkit.jl.