Peter Hill

The replacement of `DataFile` with `OptionsNetCDF` for the main dump file is almost complete. Still need to replace it in `GridFile`, but I'm hoping that's fairly self-contained and easy. A...

Yeah, `guards` does sound useful -- always including boundaries would be essential. Sounds like that could wait for another PR though :) Ok, next stumbling block: `GridFromFile`. The bulk of...

Oh, the other difficulty with option 2 is that when we read in the data in `OptionsNetCDF`, we read 1D data into an `Array`, 2D data into `Matrix`, and 3D...

`ctest` has this built-in already with the `-j` flag. We can also set the [`PROCESSORS`](https://cmake.org/cmake/help/latest/prop_test/PROCESSORS.html#prop_test:PROCESSORS) property on each test to help out here

The `PETSC_VERSION_GE` family of macros works by checking major/minor/patch. Could do something similar: ```cpp #define BOUT_VERSION_GE(MAJOR, MINOR, PATCH) \ ((BOUT_VERSION_MAJOR > MAJOR) \ || ((BOUT_VERSION_MAJOR == MAJOR) \ && ((BOUT_VERSION_MINOR...

We have something in now that deals with this, and we can deal with `x, y, z` at staggered locations. Except... I think we're still missing some bits. `FieldFactor::create*D` takes...

Oh that arithmetic performance example is maybe a little out of date now. Timings for a full model might be more instructive, like blob2d or elm-pb.

See #2207 for some discussion

Possible directions for investigation: - convergence checks for the round-trip test here: if the issue is one of mismatched discretisations, the error should decrease as we increase the resolution -...

Yes, completely onboard with this. I think ideally the output API would mirror the `Options` API for input. Ben made a start with the `OptionsNetCDF` class. I'd love for us...