Water depth per cell in metres, row-major, never negative.
Id of the domain this state belongs to.
Cumulative volume removed by declared drains since step 0, in m³.
Cumulative volume that left across open edges since step 0, in m³.
Cumulative volume admitted by declared sources since step 0, in m³.
Absolute integer step index, 0 being the authored initial state.
Absolute domain-clock second, exactly step * solver.fixedStepSeconds.
+X face velocity in metres per second at row * (columns + 1) + column for columns within [0, columns]. Face column separates cells column - 1 and column; the first and last faces are domain edges.
+Z face velocity in metres per second at row * columns + column for rows within [0, rows]. Face row separates cell rows row - 1 and row; the first and last faces are domain edges.
Water volume currently held by the lattice, in m³.
The complete conserved state of a fluid domain at one absolute step.
simulateFluidDomain replays the supplied domain from its authored state for each absolute integer step, with no retained inter-call state. A seek to n costs n lattice updates and rejects a step outside [0, solver.maxSteps]. After each update it checks depth and face velocities for finiteness; returned clock/ledger totals have no corresponding check, and the depth assignment converts a NaN intermediate to zero before that check.
Cell arrays are row-major at row * columns + column. Face arrays contain one extra face along the crossed axis, with their indexing given below.
With finite representable arithmetic, the flux ledger targets initial volume plus sourceVolume minus drainVolume and outflowVolume, up to rounding. Extreme admitted fields can overflow flux or ledger intermediates; the current checks do not prove that balance for every returned record.