ropensci/tidync

NC_CHAR policy

mdsumner opened this issue · 4 comments

mdsumner commented

I think we have to ignore these, because assumptions about dimensionality just don't make sense.

Maybe possible to say ndims = ndims - 1, but even ncdf4 has issues with this.

Also need to port this logic, and a way to pick a default variable in stars.

mdsumner commented

avoid choosing them by default:

3217782

another issue is whether they should be expanded by default with hyper_array and hyper_tibble, seems to work ok so no great danger yet

mdsumner commented

Fixed!

mdsumner commented

Also see this old GDAL issue, there's no model in GDAL for character data apparently, though it would fit a RAT set up https://trac.osgeo.org/gdal/ticket/2599