rOpenSci submission notes

[mdsumner]

• questions on activate, is it generic, separate var/grids?
• set up links to files we understand, some examples of other weird sources (see mdsumner/rasterwise)
• clarify relation to rerddap
• need a vignette
• do check, and report on exceptions
• check https://github.com/ropensci/onboarding/blob/master/packaging_guide.md
• remove all raady ness
• fix appveyor
• insert install instructions (copy anglr plan)

[mdsumner]

names of thingies

I'm considering updating the name situation like this

• hyper_filter -> filter.tidync
• hyper_slice -> hyper_array
• hyper_tibble -> as_tibble.tidync
• hyper_tbl_cube -> as_tbl_cube
• activate -> ??

It's a bit incongruous for filter, given that the syntax is special for expressions lon = lon > 10 but I think that might be ok.

Is it ok that filter.tidync returns a tidync object, rather than a tibble? If that were ok then also slice is exactly right - and we can forget hyper_slice/hyper_array.

I'm also toying with select for activating a grid or variable/s
activate I just don't know

[mdsumner]

The l3m file is not super useful as it has only 11 valid pixels, so it works fine and is small but doesn't look like it works.

[mdsumner]

Ooh, try this:

• hyper_index(1:4, 4:12, 5:8, ...) ## for raw slice versions of hyper_filter
• hyper_axes(tidync) ## to strip out the selected axis_transforms, to match hyper_slice output

Use this to construct as_tbl_cube (general) and as.RasterLayer (slice-limited) examples. The examples can make it clear that the rectilinear case fits perfectly, and even when degenerate the affine case is ignored, and the curvlinear case is just coordinates as variables, the axes are index-only. This inspired by the rerddap discussion on "melting".

[mdsumner]

I consider this ok now