Should we require that Parameters and Descriptors are given a name?

[henrikjacobsenfys]

Currently, Parameters and Descriptors are required to have a name.

However, it makes a lot of sense to create a Parameter without having to name it. For example,

a=Parameter(2.0,unit='meV')

seems like a very reasonable thing to write. We would then automatically generate a name for a. For arithmetic operations, the name of the newly generated Parameters or DescriptorNumbers is their unique id. It makes sense to do the same thing here, i.e. the expected result would be
<Parameter 'Parameter_0': 2.0000 meV, bounds=[-inf:inf]>

The problem with doing this is that name is the first argument of Parameter and DescriptorNumber. Making name optional means value becomes the first argument, and that breaks a lot. Do we want to do this?

[damskii9992]

I do think it is weird to have a mandatory name for all our parameters/descriptors. When I worked on the unique_name implementation I attempted to use the regular name as the unique_name because it made sense, unfortunately this caused a host of issues and I ended up adding the extra unique_name property instead. I would suggest that:

We remove the "name" parameter completely, I don't see the point of it.
Use the unique_name wherever the name is currently used. Possibly rename it?
Add an optional display_name or pretty_name for names to be presented in the GUI or elsewhere.
Use the unique_name in case the pretty_name or display_name is not set.

These changes WILL break a lot of stuff, so unfortunately this is no small task, despite its seemingly innocently looking simplicity. Since this task is in no way urgent or important, I will therefore claim that it should be postponed to a time when there is time for such less-important stuff.