Spatial Required updates list

[TorkelE]

Will add stuff here as I come up with them.

Internal improvements

• Currently, we loop through all edges when we figure out spatial interactions. We should create a matrix of all vertexes, make it two layers (for both directions) and loop through this one.
• Parameters tied to edges are used at the beginning, but not stored. This prevents callbacks that affect these rates. Should be fixed.
• We should cache a Matrix corresponding to the well-mixed (non-spatial) Jacobian, and update its value in each timestep (and fecth from there.
• Consider switching the order of unknowns i.e. varying space then species instead of species then space.
• Using a view of a slice of a sparse Jacobian; is this bad for performance?
• Properly utilise various matrix forms for reducing hopping rate storage for various Jump simulations.

Missing features

• Support for cartesian grids (both as a NxMxL block, and as a 3d-Matrix with Bools indicating existence of vertexes).
• support for defining a continuous domain that is then discretized as a PDE.
• Integration with structures declared for MoL. E.g use its representation for e.g. PDE solutions.
• Matrix-free jacobian calculations.
• Support for removing (non-spatial) conserved quantities in the ODESystem.
• Declaring spatial reaction as a part of a ReactionSystem.
• Spatial reactions (that are not just TransportReactions).
• Spatial reactions (transport and other) with rates depending on other spcies.
• Boundary condition stuff.

Othe stuff

• Utilise PDESolution to create a new way to interface with the output.
• Create plotting animation interface for displaying the output.
• Create documentation.

[isaacsas]

Other things I mentioned today

• consider switching the order of unknowns i.e. varying space then species instead of species then space.
• using a view of a slice of a sparse Jacobian; is this bad for performance?

[isaacsas]

We need to see how MOL is representing PDE solutions under the hood and tie into that ideally.

[TorkelE]

updated

[isaacsas]

Support accessing variables via the system lrs.A and such, and support getproperty. Then fix _symbol_to_var to work like MTK.