SolverStudio_00_04_08_24 (2012.10.10)

5 thoughts on “SolverStudio_00_04_08_24 (2012.10.10)”

1. amas008 - October 14, 2012 at 8:39 pm

   Suresh: I will look at support for dynamic ranges later this year; that is a great idea. In the mean time, you could use an Excel list which will grow the defined name range as you enter more data, or simply have blank rows at the end of your data which SolverStudio will ignore. Thanks again for the feedback, Andrew

   Reply
2. Suresh Ankolekar - October 12, 2012 at 11:53 pm

   "AMPL on NEOS" is a welcome feature. I just tested it with gurobi_ampl solver on NEOS server. It worked flawlessly and turnaround was quite fast. The results reported by the solver in the model input window matched perfectly with the ones with local Gurobi (5.0) solver. I encountered one issue. Of the three display items, namely, option display_1col 9999999; display zghj > Sheet; display fgk > Sheet; display ENPV > Sheet; # the SolverStudio wrote the first two items, zghj and fgk, (both ranges) correctly to the worksheet, but failed to write the last item (single cell), though it said in the model output window that it did so, as per the following: ... _display 0 1 1 ENPV 18718.772319343603 ## AMPL run completed on NEOS. ## Results loaded for data items: zghj* fgk* ENPV* ## (*=data item values changed on sheet) ## Done This, probably 'off by 1' kind of issue is only with the "AMPL on NEOS" language choice. The SolverStudio picks up the last item correctly with regular AMPL using local Gurobi solver with the same model. Thanks for a very useful package! Suresh

   Reply
   • amas008 - October 14, 2012 at 3:35 am

     Suresh, Thanks for the feedback and bug report. I'll look at this next week. (The NEOS parsing code is new, so it must have a bug in it!) Cheers, Andrew. PS: What are you using SolverStudio for?

     Reply
     • Suresh Ankolekar - October 14, 2012 at 6:49 am

       Andrew, Thanks for your response. I have been using SolverStudio to solve a stochastic integer programming model. I started with OpenSolver, but could not scale up. SolverStudio with AMPL using full version of Cplex and Gurobi scaled up quite well. With "AMPL on NEOS", now I hope to be able to collaborate with colleagues who do not have local access to Gurobi. I noticed one more issue, not a bug, but possibly an item for feature enhancement, if this is not an Excel issue. The SolverStudio 'Edit Data' does allow Named Ranges as Index Range(s) specifications, but converts them (together with Cell Range) to actual ranges on Update Data Item. That makes it rather impractical to use Dynamic Named Ranges (defined typically as OFFSET formula), besides loss of intuitive appeal of Named Ranges. Of course, this could possibly be an Excel issue, as I noticed that Excel inexplicably also converted the formula for the Dynamic Named Range to actual range, after I exit the Edit Data dialog. Dynamic Named Ranges would be a convenient way to handle AMPL entities (set, param, var) without having to manually redefine them in Excel Names Manager and SolverStudio Edit Data for every run. Thanks. Best regards, Suresh

       Reply
   • amas008 - October 14, 2012 at 10:41 pm

     Fixed this; vars() working differently in Python inside a subroutine (as it should) so my update was being lost. Andrew

     Reply