Merge pull request #93 from mariohsouto/improve_print

Improving print to use scientific formatting

docs/src/manual.md

@@ -35,7 +35,8 @@ A quick JuMP example:
 using ProxSDP, JuMP
 
 # Create a JuMP model using ProxSDP as the solver
-model = Model(with_optimizer(ProxSDP.Optimizer, log_verbose=true))
+model = Model(ProxSDP.Optimizer)
+set_optimizer_attribute(model, "log_verbose", true)
 
 # Create a Positive Semidefinite variable
 @variable(model, X[1:2,1:2], PSD)

examples/jump.jl

@@ -3,7 +3,10 @@ push!(Base.LOAD_PATH, path)
 using ProxSDP, JuMP
 
 # Create a JuMP model using ProxSDP as the solver
-model = Model(with_optimizer(ProxSDP.Optimizer, log_verbose=true, tol_gap = 1e-4, tol_feasibility = 1e-4))
+model = Model(ProxSDP.Optimizer)
+set_optimizer_attribute(model, "log_verbose", true)
+set_optimizer_attribute(model, "tol_gap", 1e-4)
+set_optimizer_attribute(model, "tol_feasibility", 1e-4)
 
 @variable(model, X[1:2,1:2], PSD)
 

examples/sos.jl

@@ -5,8 +5,11 @@ using DynamicPolynomials
 using ProxSDP, SCS
 
 # Using ProxSDP as the SDP solver
-model = SOSModel(with_optimizer(ProxSDP.Optimizer, log_verbose=true, max_iter=100000, full_eig_decomp=true))
-# model = SOSModel(with_optimizer(SCS.Optimizer, max_iters=100000))
+# model = SOSModel(with_optimizer(ProxSDP.Optimizer, log_verbose=true, max_iter=100000, full_eig_decomp=true))
+model = SOSModel(ProxSDP.Optimizer)
+set_optimizer_attribute(model, "log_verbose", true)
+set_optimizer_attribute(model, "max_iter", 100000)
+set_optimizer_attribute(model, "full_eig_decomp", true)
 
 @polyvar x z
 @variable(model, t)

src/printing.jl

@@ -98,19 +98,19 @@ function print_progress(residuals::Residuals, p::Params, opt, val = -1.0)
     dual_res = residuals.dual_residual[p.iter]
     s_k = Printf.@sprintf("%d", p.iter)
     s_k *= " |"
-    s_s = Printf.@sprintf("%.5f", residuals.dual_gap[p.iter])
+    s_s = Printf.@sprintf("%.2e", residuals.dual_gap[p.iter])
     s_s *= " |"
-    s_o = Printf.@sprintf("%.3f", residuals.prim_obj[p.iter])
+    s_o = Printf.@sprintf("%.2e", residuals.prim_obj[p.iter])
     s_o *= " |"
-    s_f = Printf.@sprintf("%.5f", residuals.feasibility[p.iter])
+    s_f = Printf.@sprintf("%.2e", residuals.feasibility[p.iter])
     s_f *= " |"
-    s_p = Printf.@sprintf("%.5f", primal_res)
+    s_p = Printf.@sprintf("%.2e", primal_res)
     s_p *= " |"
-    s_d = Printf.@sprintf("%.5f", dual_res)
+    s_d = Printf.@sprintf("%.2e", dual_res)
     s_d *= " |"
-    s_target_rank = Printf.@sprintf("%.0f", sum(p.target_rank))
+    s_target_rank = Printf.@sprintf("%g", sum(p.target_rank))
     s_target_rank *= " |"
-    s_time = Printf.@sprintf("%.4f", time() - p.time0)
+    s_time = Printf.@sprintf("%g", time() - p.time0)
     s_time *= " |"
     a = "|"
     a *= " "^max(0, 9 - length(s_k))

test/jump_mimo.jl

@@ -6,7 +6,7 @@ function jump_mimo(solver, seed, n; verbose = false, test = false)
 
     nvars = ProxSDP.sympackedlen(n + 1)
 
-    model = Model(with_optimizer(solver))
+    model = Model(ProxSDP.Optimizer)
     @variable(model, X[1:n+1, 1:n+1], PSD)
     for j in 1:(n+1), i in j:(n+1)
         @constraint(model, X[i, j] <=  1.0)

test/jump_randsdp.jl

@@ -2,7 +2,7 @@ function jump_randsdp(solver, seed, n, m, verbose = false)
 
     A, b, C = randsdp_data(seed, m, n)
 
-    model = Model(with_optimizer(solver))
+    model = Model(ProxSDP.Optimizer)
     @variable(model, X[1:n, 1:n], PSD)
     @objective(model, Min, sum(C[i, j] * X[i, j] for j in 1:n, i in 1:n))
     @constraint(model, ctr[k in 1:m], sum(A[k][i, j] * X[i, j] for j in 1:n, i in 1:n) == b[k])

test/jump_sdplib.jl

@@ -6,7 +6,7 @@ function jump_sdplib(solver, path; verbose = false, test = false)
     n, m, F, c = sdplib_data(path)
 
     # Build model
-    model = Model(with_optimizer(solver))
+    model = Model(ProxSDP.Optimizer)
     @variable(model, X[1:n, 1:n], PSD)
 
     # Objective function

test/jump_sensorloc.jl

@@ -3,7 +3,7 @@ function jump_sensorloc(solver, seed, n; verbose = false, test = false)
 
     m, x_true, a, d, d_bar = sensorloc_data(seed, n)
 
-    model = Model(with_optimizer(solver))
+    model = Model(ProxSDP.Optimizer)
 
     # Build SDP problem
     @variable(model, X[1:n+2, 1:n+2], PSD)