Originally posted by: JaneM

I'm try to move a model that works on my perpetual license that works on my computer to DOCoud.  I'm having problems with the following statements.  It seems that the coding for setting up sparse models does not work in the cloud version.  Is there other code that works?

Assign is an Excel file sheet that has only some of the locations.

AssignAll is the set of all PlGroups and LOC1 pairs.

 

tuple TPlGroups {

  string PGLevel2ID;

}

{TPlGroups} PlGroups = ...;

 

tuple TLOC1 {

  string LOC1;

}

{TLOC1} LOC1 = ...;

 

tuple Pairs {

  string PGLevel2ID;

  string Loc1Name;

  };

 

setof(Pairs) Assign = ...;

 

setof(Pairs) AssignAll = {<g,l> | g in PlGroups, l in LOC1};

 

 

I get the following errors.

[8/22/18, 2:09:55 PM, SEVE] 162:1-162:83 TestV_14-SimpleCloud.mod: Cannot use type {<<PGLevel2ID:string>,<LOC1:string>>} for {Pairs}.
[8/22/18, 2:09:55 PM, SEVE] 162:1-162:83 TestV_14-SimpleCloud.mod: Processing failed.

#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi

whether you rely on the cloud or not, you should change

setof(Pairs) AssignAll = {<g,l> | g in PlGroups, l in LOC1};

into

setof(Pairs) AssignAll = {<g.PGLevel2ID,l.LOC1> | g in PlGroups, l in LOC1};

regards

https://www.linkedin.com/pulse/making-decision-optimization-simple-alex-fleischer/

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

Thanks Alex, this seems to have solved this problem.  

I have a new problem now and will set up as a separate topic.

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

I was having another challenge with my setof(..) notation.  I think I solved it while writing the question.  I will leave this comment in the forum in case others are having the same problem.

All of my current challenges transferring my desktop perpetual license model to the cloud have been reflecting the same formatting which I don't understand.  In this situation, you changed

<g,l> to

<g.PGLevel2ID,l.LOC1>. 

In my other question today, the solution was to change 

{string} LOC1X =LOC1 union SLOC1_GR;

to

{string} LOC1X = {i.LOC1 | i in LOC1} union SLOC1_GR;

 

 

 

So it seems I don't understand how to use this additional location.  

Here is the next code that worked on my desktop but not in the cloud.

 

 

setof(string) LOC_PGF[g in PlGroups] =

{l | <g,l> in AssignT};

 

It looks like I solved it with the following.  However, I didn't know that one could add multiple requirements separated by comments.

setof(string) LOC_PGF[g in PlGroups] =

  {l.LOC1 | l in LOC1, <g.PGLevel2ID, l.LOC1> in AssignT};

 

Is there another better solution? 

 

 

 

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

More challenges moving to the cloud.  Is this problem similar to those above? 

My notation in this code, is giving me an error running in the cloud for probably a similar reason.  

 

dexpr float ZLWS_Occ[l in LOC1][t in TimeE] =

          sum(gl in AssignT:gl.Loc1Name==l) ZPG_WSOcc[gl][t] ;

 

The error message says that the == is wrong.

 

Based on reviewing Alex's Cours2.pdf, I tried the following, which also didn't work

 

 

dexpr float ZLWS_Occ[l in LOC1][t in TimeE] =

          sum(<g,m> in AssignT: l == m) ZPG_WSOcc[<g,m>][t] ;

          

dexpr float ZLWS_Occ[l in LOC1][t in TimeE] =

          sum(<g,l> in AssignT) ZPG_WSOcc[<g,l>][t] ;

 

 

 

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

It looks like I solved this one with:

 

dexpr float ZLWS_Occ[l in LOC1][t in TimeE] =

          sum(gl in AssignT:gl.Loc1Name==l.LOC1) ZPG_WSOcc[gl][t] ;      

 

seems as though anything in brackets needs to have the l.LOC1 or gl.Loc1Name type of format

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

Add another likely related problem.  

Earlier in my code for cloud, the following was accepted.

setof(string) LOC1F =   

  {l | <g,l> in AssignT};    

 

However, in the section to set constraints, I had the following problems. 

   ctInitializeWkSpO: forall(l in LOC1F)

            ZLWkSp[l][0] == LWSTotalT0[l] * LCtrlEx[l][0];

 

led to the error

 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

It seems to be a problem that these variables are specified over LOC1 rather than LOC1F, which is a subset of LOC1

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

Here's another one that I can't figure out that seems related to the same notation. 

dexpr float ZLWS_Use[l in LOC1][t in TimeE] =         

          sum(gl in AssignT:gl.Loc1Name==l.LOC1) (ZPG_WSOcc[gl][t] * (1 + BufferG[gl.PGLevel2ID][t])) ;

 

 

BufferG is defined as 

BufferG[PlGroups][TimeT]

I want to use BufferG for the g that is in gl

The error is: 

Cannot use type string for <PGLevel2ID:string>

 

I tried this which worked.  But I don't think I should have to use a sum when I am just trying to find the one g that equals gl.PGLeve2ID.  I'm not even sure if it is doing the right calculation.

 

dexpr float ZLWS_Use[l in LOC1][t in TimeE] =         

sum(gl in AssignT:gl.Loc1Name==l.LOC1)

(ZPG_WSOcc[gl][t] * (1 + sum(g in PlGroups: gl.PGLevel2ID == g.PGLevel2ID) BufferG[g][t])) ;

 

 

 

Similary this doesn't work

forall (g in PlGroups, t in TimeE)

   ctAssignAll:  sum(gl in AssignT:gl.PGLevel2ID==g) ZPGShLoc1[gl][t] == 1;

 

Based on an example above that worked for an expression

 

dexpr float ZLWS_Occ[l in LOC1][t in TimeE] =

          sum(gl in AssignT:gl.Loc1Name==l.LOC1) ZPG_WSOcc[gl][t] ;      

 

I tried 

forall (g in PlGroups, t in TimeE)

   ctAssignAll:  sum(gl in AssignT:gl.PGLevel2ID==g.PlGroups) ZPGShLoc1[gl][t] == 1;

 

That has the error

So tried this, which worked

forall (g in PlGroups, t in TimeE)

   ctAssignAll:  sum(gl in AssignT:gl.PGLevel2ID==g.PGLevel2ID) ZPGShLoc1[gl][t] == 1;

 

I don't understand why in one case, the index is followed by the name of the set (l.LOC1) and in this case, the name of the set doesn't work (g.PlGroups), it works when followed by the field name in the index gl,  g.PGLevel2ID.

 

Overall, I don't understand why CPLEX has set up the DOCloud version so that what works on a desktop perpetual license doesn't work in the cloud. 

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: JaneM

So I found the answer to the last question.  In my file the name of the set and the name of the variable were the same, which is why it seemed that the index was followed by the name of the set.  Once I changed the name of the variable in the set, the code is consistent with the name of the variable after the relevant index.

#DecisionOptimization
#OPLusingCPLEXOptimizer