Added functions to interact with JuMP

[caioluke]

closes #1

• Creation of necessary functions in order to interact with the JuMP interface, e.g MOI.is_empty, MOI.get, MOI.set, MOI.supports

• Creation of mutable struct 'ParameterValue' which is used to assign value to parameters

• Correction of a bug in which a ScalarQuadraticFunction was not treated as a ScalarAffineFunction when the quadratic term was a product of 2 parameters

• Added some tests with JuMP

optimizer = POI.ParametricOptimizer(GLPK.Optimizer())
    

c = [4.0, 3.0]
A1 = [2.0, 1.0, 1.0]
A2 = [1.0, 2.0, 1.0]
b1 = 4.0
b2 = 4.0

# ===== Create model with parameters =====
model = direct_model(optimizer)

@variable(model, x[i=1:2] >= 0)

# Parameters set to 0
@variable(model, y in POI.Parameter(0))
@variable(model, w in POI.Parameter(0))
@variable(model, z in POI.Parameter(0))

@constraint(model, 2*x[1] + x[2] + y <= 4)
@constraint(model, x[1] + 2*x[2] + z <= 4)

@objective(model, Max, sum(c[i]*x[i] for i=1:2) + w)

optimize!(model)
println(model)

Max 4 x[1] + 3 x[2]
Subject to
 x[1] >= 0.0
 x[2] >= 0.0
 x[2] + 2 x[1] <= 4.0
 2 x[2] + x[1] <= 4.0

# ===== Set parameter y to 2 =====
# A better way of assigning values to parameters
MOI.set(model, POI.ParameterValue(), y, 2.0)

optimize!(model)
println(model)

Max 4 x[1] + 3 x[2]
Subject to
 x[1] >= 0.0
 x[2] >= 0.0
 x[2] + 2 x[1] <= 2.0
 3 x[1] + 2 x[2] <= 4.0

[joaquimg]

@caioluke

I have an ideia to make it better.
We can make a:
struct ParameterValue <: MOI.AbstractVariableAttribute end
Then we can simply call:

MOI.set(model, POI.ParameterValue(), y, 2.0)

instead of:

cy = MOI.get(backend(model), POI.ParameterRef(), index(y))
MOI.set(backend(model), MOI.ConstraintSet(), cy, POI.Parameter(2.0))

@odow @blegat

some questions for you too...

should we add JuMP as a Test/Doc dependency and avoid as much as we can add it as an actual dependency?

Other suggestions?

[odow]

struct ParameterValue <: MOI.AbstractVariableAttribute end
Then we can simply call:

MOI.set(model, POI.ParameterValue(), y, 2.0)

👍 Yeah this would be better.

[blegat]

Looking good overall, thanks! There are however a few that are currently incorrect, I would recommend at least leaving at TODO explaining why it's incorrect

[caioluke]

@blegat with @guilhermebodin we addressed most of your comments and we left some TODO's for the future. If you have any tips on how to implement those changes that would be great.

[codecov]

Codecov Report

Merging #17 (fd38e5b) into master (d81ab0c) will decrease coverage by 3.85%.
The diff coverage is 71.42%.

@@            Coverage Diff             @@
##           master      #17      +/-   ##
==========================================
- Coverage   94.24%   90.38%   -3.86%     
==========================================
  Files           1        1              
  Lines         330      416      +86     
==========================================
+ Hits          311      376      +65     
- Misses         19       40      +21     

Impacted Files	Coverage Δ
src/ParametricOptInterface.jl	90.38% <71.42%> (-3.86%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d81ab0c...fd38e5b. Read the comment docs.