crazy dimvals

[mdsumner]

A GW file has both dimvals that are length-1 characters with all sorts (surely not intended to be used as dim-vals ever), but also one has a matrix in there.

Dealing with it like this, ignore the character ones (save for some uber future hierachical record) and flatten the matrix ones.

I've seen these character ones before, but never a matrix like this. Want to know if it's common, as it has implications for the general interpretation of curvlinear coordinates.

weird_char_dimvals <- unlist(lapply(nc$dim, function(x) is.character(x$vals)))
    dimension_values <- dplyr::bind_rows(
    lapply(nc$dim[!weird_char_dimvals], function(x) {
      ## some of these are matrices
      tibble::tibble(id = rep(x$id, length(x$vals)), vals = as.vector(x$vals))
    })
    )

ncdump::NetCDF needs a serious rework anyway, and this bogey has been hiding in the cave for a long time

[mdsumner]

This means that the vardim link will contain NAs, so either for now we have to treat those variables as unsupported. Or fix it properly.