Huite

In case this still doesn't provide enough evidence for whatever hypothesis we'd like to check, there is also the option of comparing an ad hoc implicit Euler Newton-Raphson scheme. SparseConnectivityTracer...

Another note: Newton iteration will flail hopelessly without Lipschitz continuity. Backtracking can address this problem somewhat, because if the unbounded derivative launches the state into oblivion, backtracking will detect this...

Yes, I don't think this makes sense for any explicit method. Thinking out loud: You could even go so far as to enable it exclusively for ImplicitEuler? -- Kind of...

I can confirm: VSCode Python debugger `v2024.10.0` works. For completeness: Python Debugger v2024.10.0 works with numba 0.59.1 ✔️ Python Debugger v2024.10.0 works with numba 0.60.0 ✔️ Python Debugger v2024.12.0 works...

> Ok, thanks. I'll close this PR as this doesn't seem to be Numba's fault. I'm not entirely sure about that? As I mentioned, the older numba (0.59.1) version does...

This one has github a action: https://github.com/lycheeverse/lychee/

> You are indeed cutting some corners numerically by assuming that concentrations of flows are constant over a timestep. This could be improved, but then we might get the same...

@evetion > However, and maybe @Huite can chime in, I feel that concentrations can't change that much over a timestep. They're very much related to flow itself, and the continuity...

> Do we allow for flows that would completely refresh basins (multiple times)? Yes, this doesn't have any numerical or mathematical problems for flow. Martijn and I used to run...

Are you sure this is unexpected? The chunked version will not load any data variables from disk into memory, whereas the unchunked version will. So we're looking at either loading...