Lh/extension refactor

[hellemo]

Support extending standards JuMP macros (#19):

m = Model()
@variable(m, x[i = 1:3, j = 100:102] >= 0, container = IndexedVarArray)

Should also fix #22 for IndexedVarArray

This PR includes changes from PR #24, can separate later if we don't want those changes, but is using e.g. nnz here.

[codecov-commenter]

Codecov Report

Base: 83.94% // Head: 84.26% // Increases project coverage by +0.32% 🎉

Coverage data is based on head (6c7e0e1) compared to base (5269ddd).
Patch coverage: 92.85% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #25      +/-   ##
==========================================
+ Coverage   83.94%   84.26%   +0.32%     
==========================================
  Files           8        8              
  Lines         436      445       +9     
==========================================
+ Hits          366      375       +9     
  Misses         70       70              

Impacted Files	Coverage Δ
src/sparsearray.jl	52.08% <50.00%> (ø)
src/indexedarray.jl	91.07% <100.00%> (+0.78%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[trulsf]

Nice solution to get integration with JuMP macros.

A question regarding the index sets. Currently, these are not copied upon construction so they can be modified during the lifetime of the variable. E.g. the following example will throw an exception

F = ["A", "B"]
G = [1,2,3]

@variable(m, x[F,G] >= 0, container = IndexedVarArray)

G = []
insertvar!(x, "A", 1)

Should it be an option to fix the index sets upon construction? I.e. let the index_names tuple have copies of the original index sets. I guess it is not that easy to provide an option to the @variable macro, so maybe a parametric type with the behaviour as parameter?

[hellemo]

Waiting for a decision on order of arguments (see jump-dev/JuMP.jl#3088) before merging.

[trulsf]

I think there is a problem with the linear indexing when integers are used for the index sets

@variable(m, x[10:20]; container=IndexedVarArray)
x[1]

will lead to an infinite loop.

[trulsf]

I also have some problems with sum(x) and collect(x) where x is an IndexedVarArray

[trulsf]

I am leaning towards looking more to SparseAxisArray and not so much to SparseArrays for how to implement length() and exclude size().

[hellemo]

Allright, I'll have a look, maybe revert the AbstractArray stuff if it only causes problems.