Introducing Env #443

guimarqu · 2021-02-03T09:35:16Z

mandatory argument for the algorithms
remove all global mutable variables (!)

In the future, Env will allows us to simplify algorithm architecture by moving storage in it

codecov · 2021-02-03T09:42:41Z

Codecov Report

Merging #443 (c89bb81) into master (70faf97) will increase coverage by 0.00%.
The diff coverage is 92.00%.

@@           Coverage Diff           @@
##           master     #443   +/-   ##
=======================================
  Coverage   80.13%   80.14%           
=======================================
  Files          52       50    -2     
  Lines        4299     4301    +2     
=======================================
+ Hits         3445     3447    +2     
  Misses        854      854

Impacted Files	Coverage Δ
src/Algorithm/basic/solvelpform.jl	`93.87% <ø> (ø)`
src/Algorithm/branching/branchingrule.jl	`20.00% <ø> (ø)`
src/Algorithm/branching/varbranching.jl	`97.22% <ø> (ø)`
src/Algorithm/divide.jl	`54.54% <0.00%> (ø)`
src/Algorithm/interface.jl	`88.57% <0.00%> (ø)`
src/Algorithm/preprocessing.jl	`75.44% <ø> (ø)`
src/MOIcallbacks.jl	`93.61% <ø> (ø)`
src/MathProg/MathProg.jl	`100.00% <ø> (ø)`
src/MathProg/clone.jl	`100.00% <ø> (ø)`
src/MathProg/vcids.jl	`53.57% <ø> (ø)`
... and 13 more

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 70faf97...9e85194. Read the comment docs.

vitornesello · 2021-02-03T09:48:23Z

src/Algorithm/benders.jl

@@ -51,14 +51,12 @@ end

 getoptstate(data::BendersCutGenRuntimeData) = data.optstate

-function run!(algo::BendersCutGeneration, rfdata::ReformData, input::OptimizationInput)::OptimizationOutput    
-
+function run!(algo::BendersCutGeneration, env::Env, rfdata::ReformData, input::OptimizationInput)::OptimizationOutput    


vitornesello · 2021-02-03T09:49:04Z

src/Algorithm/benders.jl

    elapsed_time = @elapsed begin
-        optresult = TO.@timeit Coluna._to "relaxed master" run!(SolveLpForm(get_dual_solution = true), ModelData(master), OptimizationInput(OptimizationState(master)))
+        optresult = TO.@timeit Coluna._to "relaxed master" run!(SolveLpForm(get_dual_solution = true), env, ModelData(master), OptimizationInput(OptimizationState(master)))


vitornesello · 2021-02-03T09:51:18Z

src/Algorithm/colgen.jl

@@ -721,7 +725,10 @@ function cg_main_loop!(
            end
        end

-        print_colgen_statistics(phase, iteration, stabstorage.curalpha, cg_optstate, nb_new_columns, rm_time, sp_time)
+        print_colgen_statistics(
+            env, phase, iteration, stabstorage.curalpha, cg_optstate, nb_new_columns, 


trailing whitespace?

vitornesello · 2021-02-03T09:51:29Z

src/Algorithm/colgen.jl

            sp_time += @elapsed begin
-                nb_new_col = solve_sps_to_gencols!(spinfos, algo, phase, data, redcostsvec, lp_dual_sol, smooth_dual_sol)
+                nb_new_col = solve_sps_to_gencols!(
+                    spinfos, algo, env, phase, data, redcostsvec, lp_dual_sol, 


trailing whitespace?

vitornesello · 2021-02-03T09:52:32Z

src/Algorithm/conquer.jl

+    node::Node, algo::AbstractConquerAlgorithm, env::Env, data::ReformData, 
+    storages_to_restore::StoragesUsageDict, opt_rtol::Float64 = Coluna.DEF_OPTIMALITY_RTOL, 


trailing whitespace?

vitornesello · 2021-02-03T09:59:16Z

src/decomposition.jl

 )
    !haskey(annotations.constrs_per_ann, mast_ann) && return
    constrs = annotations.constrs_per_ann[mast_ann]
    for (id, constr) in constrs
        cloneconstr!(
-            origform, masterform, masterform, constr, MasterMixedConstr, loc_art_var = true
+            origform, masterform, masterform, constr, MasterMixedConstr, 


trailing whitespace?

vitornesello · 2021-02-03T09:59:26Z

src/decomposition.jl

    annotations::Annotations,
-    mast_ann
+    mast_ann,


extra comma

vitornesello · 2021-02-03T09:59:44Z

src/decomposition.jl

@@ -136,7 +139,8 @@ function create_side_vars_constrs!(
        name = string("sp_lb_", spuid)
        lb_conv_constr = setconstr!(
            masterform, name, MasterConvexityConstr;
-            rhs = lb_mult, kind = Essential, sense = Greater, inc_val = 100.0, loc_art_var = true
+            rhs = lb_mult, kind = Essential, sense = Greater, inc_val = 100.0, 


trailing whitespace?

vitornesello · 2021-02-03T09:59:51Z

src/decomposition.jl

@@ -145,16 +149,17 @@ function create_side_vars_constrs!(
        name = string("sp_ub_", spuid)
        ub_conv_constr = setconstr!(
            masterform, name, MasterConvexityConstr; rhs = ub_mult,
-            kind = Essential, sense = Less, inc_val = 100.0, loc_art_var = true
+            kind = Essential, sense = Less, inc_val = 100.0, 


trailing whitespace?

vitornesello · 2021-02-03T10:00:38Z

src/decomposition.jl

-    prob::Problem, annotations::Annotations, reform::Reformulation, parent,
-    node::BD.Root
+    prob::Problem, reform::Reformulation, env::Env, annotations::Annotations, parent,
+    node::BD.Root,


extra comma

rrsadykov · 2021-02-03T11:04:08Z

I do not get the point of introducing environment. Is it just to store parameters? Storages are not global, they "belong" to a model, i.e. to the master, to a subproblem.

We need to create a document explaining the architecture of Coluna so that the changes to architecture are clear. Where is a place for such a document?

For the moment, we have models and algorithms. A model contain user data (formulation, etc) and computed data (storage). Algorithms are immutable and contain just the parameters, and they know child algorithms. I think it is good to have parameters local to algorithms. A user can pass either global parameters, which then are "localized" before solving (high-level usage). A user can also "fine-tune" local parameters of every algorithm, as it is done now (low-level usage). What is the advantage of having global parameters in an environment?

guimarqu · 2021-02-03T11:31:24Z

We need to create a document explaining the architecture of Coluna so that the changes to architecture are clear. Where is a place for such a document?

I'm writing the document, I'll send you it when I finish the first version.
I don't plan to change storages/algorithm architecture before.

What is the advantage of having global parameters in an environment?

The main goal is to remove all global mutable variables such as _params_ or _globals_.

guimarqu added 3 commits February 3, 2021 10:09

no more global mutable vars

b4c483e

rm globals

761a9bc

fix time elapsed

4bb94d2

guimarqu requested review from vitornesello and rrsadykov February 3, 2021 09:35

vitornesello reviewed Feb 3, 2021

View reviewed changes

address vitor comments

9e85194

guimarqu requested a review from vitornesello February 3, 2021 12:14

vitornesello approved these changes Feb 3, 2021

View reviewed changes

guimarqu merged commit 22537e5 into master Feb 4, 2021

guimarqu deleted the gm/env branch February 4, 2021 16:47

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Introducing Env #443

Introducing Env #443

guimarqu commented Feb 3, 2021

codecov bot commented Feb 3, 2021 •

edited

Loading

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

vitornesello Feb 3, 2021

rrsadykov commented Feb 3, 2021

guimarqu commented Feb 3, 2021 •

edited

Loading

		node::Node, algo::AbstractConquerAlgorithm, env::Env, data::ReformData,
		storages_to_restore::StoragesUsageDict, opt_rtol::Float64 = Coluna.DEF_OPTIMALITY_RTOL,

Introducing Env #443

Introducing Env #443

Conversation

guimarqu commented Feb 3, 2021

codecov bot commented Feb 3, 2021 • edited Loading

Codecov Report

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rrsadykov commented Feb 3, 2021

guimarqu commented Feb 3, 2021 • edited Loading

codecov bot commented Feb 3, 2021 •

edited

Loading

guimarqu commented Feb 3, 2021 •

edited

Loading