Tomás Chor

> Great test! I requested a change to `construct_output` so that it works `with_halos`. @glwagner did you catch my comment on your suggestion? Is it better to try to make...

Posting this here for my own reference later. With the current code, when I run this example: ``` using Oceananigans using Oceananigans.OutputWriters: construct_output N = 4 grid = RectilinearGrid(topology =...

> @tomchor you need to generalize this line: > > https://github.com/CliMA/Oceananigans.jl/blob/0c23eeaca4f5d1888c3d78f85de3ff221a732bb4/src/Fields/field.jl#L225 > > to work when `parent(data)` is a `SubArray`. I think here we want a recursive parent that extracts...

Actually, @glwagner, your code worked out of the box :)

> @tomchor I found some dispatch on `ComputedField` --- but nothing about the output construction should depend on whether a field has an abstract operation underneath or not; all of...

> I guess the simplest solution is just to ignore "output writer indices" for fields that have non-default indices. And this is the first solution discussed on this PR... but...

> Ok, let me know what you think. Looks good! From my local tests it seems like even these tests are passing: https://github.com/CliMA/Oceananigans.jl/blob/87f8093477104f112c107de1b3f0e677aa2498d1/test/test_output_writers.jl#L56-L57 Although I thought they had to fail...

Test are passing! @glwagner please lmk if you want changes in the test. If not, I'll merge and register a new version.

Just an update (mostly for my own benefit later), apparently the outputs are constructed appropriately now, but the outputs still can't be written to NetCDF because the writer creates the...

@navidcy I'm pretty sure https://github.com/CliMA/Oceananigans.jl/pull/2502 fixed at least part of the failing tests here. I'm not sure how to deal with some of the conflicts so I'll leave the merging...