Adriano Fantini

Manually processing different chunks is an option and it works well, but I was wondering if some automated procedure can be applied. In general however I do not understand why...

I usually do this with `ncdf4` and never encountered problems.

Weird. I suspect that `_ChunkSizes` is NA in `ncdf4` since the chunk size is equal to the variable size - that is, there is only one chunk in the file.

> Another point on this is that the variable 'pp' is internally a short integer, but gets expanded into double by the "unpacking" (application of scale/offset). In tidync that is...

> In an ideal world, I think, one would entirely get away from `proxy=FALSE` You mean `TRUE`? >unless the object is created in-memory, but even then you could write it...

> @adrfantini -- the chunking of time being most common, IMHO, is a reflection of time being the most common default outer dimension (required in COARDS?). In NetCDF3, this just...

> Yes netcdf-dev, there is a new read_stars_tidync function. Units for dins and vars, and consolidation of the two funs is coming, was a good day :) @mdsumner `mdsumner/stars@netcdf-dev`? Because...

> oh sorry, it's only internal - I didn't think of that. It's still too early unless you like ~suffering~ chasing development - `devtools::load_all()` is the fastest way to load...

>  heuristics to determine that 2D vars are curvilinear coords of 3D+ vars If a netCDF file follows the CF-Conventions, this should be indicated by the `coordinates` attribute. If...

#12 and #54 contain links to properly CF-Compliant files with 2, 3 and 4D variables, with spatial coordinates in LCC projection, which is the EURO-CORDEX grid we usually use. Do...