Consistently map function names to their "canonical" names

[ranocha]

In GitLab by @sloede on May 28, 2020, 08:35

As a follow up to a discussion with @ranocha, I think we should use a consistent way to map function names to their canonical names for documentation/parameter file purposes.

Example

The function initial_conditions_gauss_pulse should be consistently be called either initial_conditions_gauss_pulse or just gauss_pulse when referring to it in strings. That is, we want to either use

initial_conditions = "initial_conditions_gauss_pulse"

and have

julia> get_name(initial_conditions_gauss_pulse)
"initial_conditions_gauss_pulse"

or

initial_conditions = "gauss_pulse"

and have

julia> get_name(initial_conditions_gauss_pulse)
"gauss_pulse"

Since for me, the initial_conditions_ prefix is actual not part of the function name but is used to put it in a namespace, I prefer the second variant, as context alone should be sufficient to determine what kind of function it is. It reduces typing and feels more natural to me, similarly, I wouldn't say "I am Human Michael" but just "I am Michael" (unless I need to emphasize that I am not your machine overlord).

[ranocha]

In GitLab by @sloede on May 28, 2020, 08:36

mentioned in merge request !57

[ranocha]

In GitLab by @ranocha on May 28, 2020, 08:49

I prefer the second variant, as context alone should be sufficient to determine what kind of function it is.

We can do that. The required changes seem to be

1. Specialize get_names accordingly.
2. Add additional logic to the parsing part, e.g.

   julia> fluxname = "chandrashekar"
   "chandrashekar"
   
   julia> try 
              flux = eval(Symbol(fluxname))
          catch e
              flux = eval(Symbol(fluxname * "_flux"))
          end

[ranocha]

In GitLab by @sloede on May 28, 2020, 08:55

Ah, I forgot that my approach would also require external users to always use a _flux subscript for their new flux functions, which is of course very inconvenient and impractical. OTOH, allowing both forms in a parameter file, e.g., central and central_flux is also suboptimal - there should always be just a single name.

Maybe we just use the long names for now and bench this discussion until we have settled on an approach for how to handle parameters in the future?

[ranocha]

In GitLab by @sloede on May 28, 2020, 09:43

changed title from Consistently map function names to {-its-} "canonical" names to Consistently map function names to {+their+} "canonical" names

[ranocha]

In GitLab by @sloede on May 28, 2020, 10:57

changed the description

[ranocha]

In GitLab by @gregorgassner on May 28, 2020, 11:28

In my personal opinion, the current version as in the merge request is fine for me! I do not mind to type initial_condition one more time.

Imo, it makes it clear to read and that is good in my book.