Migration guide (to 0.1)

Version 0.1 reshapes the internals for automatic differentiation and future neural-CA support. This page lists what changed and how to update calling code.

Field access replaced with accessor functions

Struct fields are no longer public API; use the accessor functions instead.

Before	Now
`automaton.evolution`	`evolution_history(automaton)`
`automaton.rule`	`cellular_automaton_rule(automaton)`
`automaton.generations`	`generation_count(automaton)`
`dca.ruleset`	`rule_lookup_table(dca)`
`dca.radius`	`neighborhood_radius(dca)`
`dca.states`	`cell_state_count(dca)`

Discrete rule states must have an `Integer` eltype

DCA/TCA now validate the state array and reject non-integer eltypes, even when the values are integer-valued.

Before	Now
`zeros(ncells)`	`zeros(Int, ncells)`

Abstract type hierarchy renamed

Only relevant if you dispatched on these types directly.

Before	Now
`AbstractCA`	`AbstractCellularAutomaton`
`AbstractRule`	`AbstractCellularAutomatonRule`
`AbstractODRule`	dimension/kind-specific subtype (see below)
`AbstractTDRule`	dimension/kind-specific subtype (see below)
`AbstractDCARule`	`AbstractDiscreteCellularAutomatonRule`
`AbstractCCARule`	`AbstractContinuousCellularAutomatonRule`
`AbstractTCARule`	`AbstractTotalisticCellularAutomatonRule`
`AbstractLifeRule`	`AbstractLifeLikeCellularAutomatonRule`

New functional core: `next_state`/`rollout`

The callable interface (dca(state), life(state), ...) still works, but now delegates to next_state/rollout, which are the recommended entry points going forward and required for AD use.

Before	Now
`dca(state)`	`next_state(dca, state)` (or keep calling `dca(state)`)
`CellularAutomaton(rule, ic, n)` loop	`rollout(rule, ic, n; save=true)` for a bare trajectory

Boundary conditions are now explicit

Previously periodic wraparound was hardcoded. It is now the Periodic() default of a new AbstractBoundaryCondition family, passed via the boundary keyword. Default behavior is unchanged unless you relied on internal helpers.

Before	Now
(hardcoded periodic)	`next_state(rule, state; boundary=Periodic())`
	`Reflecting()` and `ConstantBoundary(value)` also available