Jet substructure structs - do we need them?

[graeme-a-stewart]

Discussing with @m-fila he pointed out that the current scheme of calling mass_drop, soft_drop, jet_filtering and jet_trimming with special structs is sub-optimal. These structs simply gather together a few parameters and pass them into the function. There's no great value in having them as structs, it seems a bit overkill for functions with a few parameters.

A better option is to convert the functions to take named parameters, e.g.,

mass_drop(jet::PseudoJet, clusterseq::ClusterSequence; mu, y)

If the user wants to gather the parameters together, for consistency, then this can be done as Julia named tuple and an argument splat:

md_args = (mu=0.67, y=0.09)
mass_drop(jet, clusterseq; md_args...)

For soft drop, where there is a sensible default for the cluster radius then this can have the normal argument default:

function soft_drop(jet::PseudoJet, clusterseq::ClusterSequence; zcut, beta, radius=1.0)

N.B. I would propose that we rename the second parameter beta (β is probably a bit too far...).

Thoughts on this @m-fila @sattwamo?

[m-fila]

Just for bookkeeping, I'm not sure what is the benefit in keeping them as structs. One thing I can think of is that maybe a struct based looks more similar to FastJet?
To me having them as keywords arguments is cleaner and more natural while the usability stays the same.

N.B. I would propose that we rename the second parameter beta (β is probably a bit too far...).

Probably I'm not fully converted yet but I don't like the idea of having fancy glyphs as keyword arguments. But beta is fine

I'm not sure how far away is 0.5.0 release, but probably it's possible even now to add functions with keyword arguments and deprecate the functions with structs?

[sattwamo]

Just as @m-fila stated, I had used the structs to make an interface similar to FastJet. But I agree that using named parameters might be better in this case. I can make the necessary changes and create a new PR.

[graeme-a-stewart]

Yes, @m-fila - this is exactly what I thought. And the trick of splatting a named tuple for named parameters is very idiomatic Julia.

Thanks for agreeing to handle this, @sattwamo.