NLsolve.jl Types as API for choosing methods

I do think that we need to adapt this here, just as we did in Optim. Is anyone strongly against?

Apr 24 '17 19:04 pkofod

I think this is a good idea. Then it is easy for algorithms to have their own set of algorithm-specific parameters that are easy for users to set.

Apr 24 '17 19:04 sglyon

I think this is a good idea. Then it is easy for algorithms to have their own set of algorithm-specific parameters that are easy for users to set.

:+1:

Apr 24 '17 19:04 pkofod

Alright, so a tiny problem here.. Do we call them Newton and NewtonTrustRegion as in Optim? If we don't, it''ll seem weird that they have different names. If we do, you need to prefix NLsolve. or Optim. if both are loaded. Does that seem like a reasonable thing do have? That you need to qualify if both are loaded.

Jun 22 '17 05:06 pkofod

I think Newton and NewtonTrustRegion are good suggestions.

Also prefixing with the module name is totally reasonable if both are loaded

Jun 26 '17 21:06 sglyon

Great, I'll have a stab at it one of these days.

Jun 27 '17 05:06 pkofod

Is this still on the todo list? If so, I might be able to have a stab at it. It'd be quite useful for me in a particular use case that I've got where I'd like to be able to choose a solver in a higher level interface and pass down method specific parameters without special casing the methods.

Jun 17 '18 22:06 dawbarton

Yes it is. I think I have a branch somewhere that's half way there, but probably against an old master, so feel free to give it a go!

Jun 18 '18 05:06 pkofod