Closed abelsiqueira closed 4 years ago
One disadvantage is that NLPModels.jl
will not be able to define any easy models, only models that define all needed functions, such as HS5
. We could keep ADNLPModels
to prevent that.
After discussing it, we decided not to split anymore.
I'm trying to rebase PR #187 and other 6 parts of #166 (Complete update of NLPModels.jl) and I'm finding very difficult to update everything because of the large amount of code. This happens on a smaller scale, for instance, when we decide to create a new function, which needs to be defined for every single model.
I don't remember the reasons to join
NLSModels.jl
here, but know I'm thinking that maybe we should split a few packages for easier maintainability. Considering that releasing packages is a lot easier know, this seems feasible by always keeping lower and upper bounds on the versions. Secondary less used packages can be updated in a later date.NLPModelsJuMP.jl and CUTEst.jl are examples of how things will be after the change.
Parts suggestion:
@dpo, what do you think?