Improving get_field of nomenclature module

[leifdenby]

Just copying over your comments @LSaffin from #30 (comment)

The changes I would make are

1. I would put the unit conversion in a separate function and module because it doesn't cover nomenclature
2. When using get_field in other functions it would be better to use it from eurec4a_environment rather than eurec4a_environment.nomenclature
3. Considering the first two points, it would be good to change the name of the function in nomenclature (e.g. to get_field_by_name) and have the function get_field in the top level module call that function and the function for unit conversion. Then other functions are calling this top level get_field and you can add other functionality that might not be related to nomenclature to it.

I can make these changes once your pull request is merged if they sound OK to you. Other questions I had:

1. Did you find any important differences between cf_units and cfunits? As far as I could tell they were basically the same but cf_units was more active and better documented.

2. Was using iris considered at all? This already does the unit checking as well as coordinate checking and better support for CF standard names (although the CF standard naming could be a limitation).

[leifdenby]

I would put the unit conversion in a separate function and module because it doesn't cover nomenclature

This sounds good @LSaffin, yes good idea

2. When using `get_field` in other functions it would be better to use it from `eurec4a_environment` rather than `eurec4a_environment.nomenclature`

Ok, I don't have a strong preference on this. If you think that's better then let's do that.

3. Considering the first two points, it would be good to change the name of the function in nomenclature (e.g. to `get_field_by_name`) and have the function `get_field` in the top level module call that function and the function for unit conversion. Then other functions are calling this top level `get_field` and you can add other functionality that might not be related to nomenclature to it.

Yes, that sounds alright. I would like to be able to keep calling get_field(ds=ds, field_name=nom.TEMPERATURE, units='K' though, so that units keeps being an argument. Is that ok with you? Just seems a bit of a hassle to be calling two functions each time.

I can make these changes once your pull request is merged if they sound OK to you. Other questions I had:

Great!

1. Did you find any important differences between cf_units and cfunits? As far as I could tell they were basically the same but cf_units was more active and better documented.

Only that cfunits failed to install unless the UDUNITS system library is installed. In some cases we might not need to convert a unit and so I thought it would be nice to only require the system library to be there when needed. But maybe this is a bad idea and we should just enforce that people have UDUNITS installed on their system. We could also use conda instead of pip requirements, pip is just a bit more light-weight. We could change to cfunits in that case.

2. Was using iris considered at all? This already does the unit checking as well as coordinate checking and better support for CF standard names (although the CF standard naming could be a limitation).

I thought iris was a bit overkill here since it is quite a big install. Maybe if we find datasets we can't support with just cf_units we make iris a requirement?

[leosaffin]

That all sounds good. I'll put together a pull request with the changes for you to approve.

That sounds like a good reason to go with cf_units. I won't worry about iris either, just making sure we don't end up remaking that library.

[leifdenby]

That sounds like a good reason to go with cf_units. I won't worry about iris either, just making sure we don't end up remaking that library.

Yes, definitely. We might want to start using it if we find we need more functionality that iris gives. I'll close this now since I've just merged #33. Thanks again for this