add no_start_opt_check + opt0 not a vector

src/Stopping/GenericStoppingmod.jl

@@ -115,18 +115,19 @@ end
 """
  start!: update the Stopping and return a boolean true if we must stop.
 
- `start!(:: AbstractStopping; kwargs...)`
+ `start!(:: AbstractStopping; no_start_opt_check :: Bool = false, kwargs...)`
 
  Purpose is to know if there is a need to even perform an optimization algorithm
- or if we are at an optimal solution from the beginning.
+ or if we are at an optimal solution from the beginning. Set *no\\_start\\_opt\\_check*
+ to *true* to avoid checking optimality.
 
  The function *start!* successively calls: *\\_domain\\_check*, *\\_optimality\\_check*,
  *\\_null\\_test*
 
  Note: - *start!* initialize the start\\_time (if not done before) and *meta.optimality0*.
        - Keywords argument are sent to the *\\_optimality\\_check!* call.
 """
-function start!(stp :: AbstractStopping; kwargs...)
+function start!(stp :: AbstractStopping; no_start_opt_check :: Bool = false, kwargs...)
 
  stt_at_x = stp.current_state
  x        = stt_at_x.x
@@ -141,11 +142,11 @@ function start!(stp :: AbstractStopping; kwargs...)
  end
 
  stp.meta.domainerror = _domain_check(stp.current_state)
- if !stp.meta.domainerror
+ if !stp.meta.domainerror && !no_start_opt_check
    # Optimality check
    optimality0          = _optimality_check(stp; kwargs...)
-   stp.meta.optimality0 = optimality0
-   if isnan(optimality0)
+   stp.meta.optimality0 = norm(optimality0, Inf)
+   if (true in isnan.(optimality0))
      stp.meta.domainerror = true
    end
 

test/examples/run-optimsolver.jl

@@ -116,8 +116,7 @@ include("activeset.jl")
 
 printstyled("Constrained optimization: active-set algorithm tutorial.\n", color = :green)
 x0 = 1.5*ones(6);x0[6]=1.0
-nlp_bnd = ADNLPModel(rosenbrock,  x0,
-                 lvar = fill(-10.0,size(x0)), uvar = fill(1.5,size(x0)))
+nlp_bnd = ADNLPModel(rosenbrock,  x0, fill(-10.0,size(x0)), fill(1.5,size(x0)))
 
 nlp_bnd_at_x = NLPAtX(x0)
 stop_nlp_c = NLPStopping(nlp_bnd, max_iter = 10)
@@ -141,9 +140,10 @@ include("penalty.jl")
 printstyled("Constrained optimization: quadratic penalty tutorial.\n", color = :green)
 x0 = 1.5*ones(6)
 c(x) = [sum(x)]
+# ADNLPModel(f, x0, lvar, uvar, c, lcon, ucon)
 nlp2 = ADNLPModel(rosenbrock,  x0,
-                 lvar = fill(-10.0,size(x0)), uvar = fill(10.0,size(x0)),
-                 y0 = [0.0], c = c, lcon = [-Inf], ucon = [5.])
+                  fill(-10.0,size(x0)), fill(10.0,size(x0)),
+                   c, [-Inf], [5.])
 
 nlp_at_x_c = NLPAtX(x0, zeros(nlp2.meta.ncon))
 stop_nlp_c = NLPStopping(nlp2, nlp_at_x_c, atol = 1e-3,