Gribberish name in xarray backend is difficult to parse/generate

[henrymartin1]

Hi,
really awesome library! I was working on ingesting a subset of GFS data and gribberish is really quite handy with that. I was very happy about the xarray backend that is capable of ingesting the whole gribfile as array.
However, I think that the given to the variables in xarray is currently difficult to work with. I have two issues at the moment:

• Variable name changes depending on total dataset ingested:
  If a base variable name exists only once in a grib file, it is named as only the basename (e.g., temp) but if the basename exists multiple times than the same variable becomes basename + hash (e.g., tmp_TMPsigma_fcst. I think it would be easier to work with if the variable name would always be basename + hash no matter what.

• some parts of the hash have no separator:
  Currently the hash is generated as:
  {variable_name}_{variable_name.upper()}{surface_type}_{statistical_process}{generating_process}
  It therefore has two parts which are not separated by an underscore (i.e., in tcdc_TCDCmcl_avgfcst). In order to split the name in its components, a fairly complex logic is required. Would it be possible to change the naming into
  {variable_name}_{variable_name.upper()}_{surface_type}_{statistical_process}_{generating_process} ?

[mpiannucci]

This is great input thank you! I think to solve this, i will split out the var, surface, stat process and the generating process to start, and also add an option to shrink the naming or not based on what has been filtered out.

I am also trying to figure out if the way cfgrib works for this is preferred, where you choose the levels you care about so you dont have these weird names. This is where i am leaning toward if that is useful

[mpiannucci]

I am also trying to figure out if the way cfgrib works for this is preferred, where you choose the levels you care about so you dont have these weird names. This is where i am leaning toward if that is useful