September release of or-tools

By -
I just released or-tools version 6.4.

The biggest change is that I reverted the ortools-<linux distro> specific pypi modules as we now have manylinux distro agnostic wheel files.

Here is the full change log:

Change in platforms

  • Pypi modules on Linux platforms are now delivered as wheel files using the manylinux1 tag. Thanks to Federico Ficarelli. With this change, we have backtracked the per-linux modules introduced in the July 2017 release.

New features

  • Improved scaling method used inside GLOP.
  • Fix wrapping of evaluators in the C# routing library. Thanks to DevNamedZed.
  • Improve the performance of the flatzinc presolve for large models.
  • Use the SAT backed for flatzinc by default.
  • Improve performance of the Core based approach for the sat solver.
  • Fix bug in the linear assignment algorithm that was failing incorrectly.
  • Added F# examples in ortools/examples/fsharp. (Thanks to Matthew Moore).
  • Remove check for positive penalties in the routing library.

Comments

  1. UnknownOctober 4, 2017 at 8:43 AM

    Hi Perron, good job with or-tools.
    I have a question, is there a way to visualize the routes in the form of lines connecting the nodes whereby the nodes are labelled with the corresponding node numbers. Of course with each route having different color.

    Thank you
    -
    Anthony

    ReplyDelete
  2. UnknownNovember 10, 2017 at 9:44 AM

    Hi Laurent,

    I am new to Or-tools as well as python (I am using anaconda python 3.6), I am try to solve 20 Job-20 Machine scheduling problem, I am using the following benchmark data: http://mistic.heig-vd.ch/taillard/problemes.dir/ordonnancement.dir/jobshop.dir/tai20_20.txt

    I have many problems, one of them when i try to solve the problem of more than 10 M-10J, I have left python more than 12 hrs, and still overloaded.

    once again, when I have tried 15-M15J, I have the following error

    IndexError Traceback (most recent call last)
    in ()
    137
    138 if __name__ == '__main__':
    --> 139 main()

    in main()
    46 horizon = 0
    47 for i in all_jobs:
    ---> 48 horizon += sum(processing_times[i])
    49 # Creates jobs.
    50 all_tasks = {}

    IndexError: list index out of range

    ReplyDelete
  3. UnknownJanuary 5, 2019 at 2:00 AM

    Which algorithm does OR-Tools use to solve Vehicle routing problems ?

    ReplyDelete
  4. UnknownJanuary 5, 2019 at 8:26 PM

    Hello Dr Laurent Perron:
    Merry Christmas and happy new year!
    　 I have concerned with operation research problems like vertex separation, minimum cost maximum flow for a long time. Currently I am doing research for supply chain financing project which provide services of factoring, forfaiting etc to customers, and I plan to design bipartite match engines between trade products to customers for each production stage, and another minimum cost maximum flow optimization solver to make most use of working capital in global view. I am very interested in Google's OR Tools which provided with some examples, and I believe we could carry out many valuable work under OR platform, so I would sincerely mail to Dr Laurent Perron to establish project connection.
    The supply chain financing project is working for a famous bank and will continue to invest and operate, I had admired OR abilities of French scientist for a long time, and I believe Dr Laurent Perron's help will be the rainbow that leads to success. I plan to visit Netherlands from Feb 13th,2019 to Feb 16th for Neo4j's graph modeling meeting. I am not sure if you still live in France or not, if you are still in Paris, I really hope to interview with Dr Laurent during this trip; if you are in Google Inc's Mountain View working place, I would plan to visit there some time in the near future.
    All in all , I am looking forward to hearing from you.
    My mail is : 1254209774@qq.com

    Best Regards
    Francis Huang
    Jan 5th, 2019
    Tongji University, Shanghai

    ReplyDelete
  5. UnknownJanuary 5, 2019 at 8:27 PM

    Hello Dr Laurent Perron:

    Merry Christmas and happy new year!

    　 I have concerned with operation research problems like vertex separation, minimum cost maximum flow for a long time. Currently I am doing research for supply chain financing project which provide services of factoring, forfaiting etc to customers, and I plan to design bipartite match engines between trade products to customers for each production stage, and another minimum cost maximum flow optimization solver to make most use of working capital in global view. I am very interested in Google's OR Tools which provided with some examples, and I believe we could carry out many valuable work under OR platform, so I would sincerely mail to Dr Laurent Perron to establish project connection.

    The supply chain financing project is working for a famous bank and will continue to invest and operate, I had admired OR abilities of French scientist for a long time, and I believe Dr Laurent Perron's help will be the rainbow that leads to success. I plan to visit Netherlands from Feb 13th,2019 to Feb 16th for Neo4j's graph modeling meeting. I am not sure if you still live in France or not, if you are still in Paris, I really hope to interview with Dr Laurent during this trip; if you are in Google Inc's Mountain View working place, I would plan to visit there some time in the near future.

    All in all , I am looking forward to hearing from you.
    My mail is : 1254209774@qq.com

    Best Regards
    Francis Huang
    Jan 5th, 2019
    Tongji University, Shanghai

    ReplyDelete

Post a Comment

Popular posts from this blog

Linking glpk with or-tools on unix

By -
Following the change on SCIP, here is the new procedure to compile and link GLPK with or-tools This currently applies to unix only: First download the latest glpk archive (4.61 at the time of this post). Untar, cd into the glpk directory on linux:   ./configure --prefix=<install_path> --with-pic   make   make install on mac OS X:   ./configure --prefix=<install_path>   make CFLAGS='--mmacosx-version-min=10.8'   make install   then add:   UNIX_GLPK_DIR=<install_path> in or-tools/Makefile.local recompile or-tools:   make clean   make cc test it:   run bin/integer_programming and bin/linear_programming. Both must display glpk runs. I have no easy solution for glpk on windows currently, still investigating.
Read more

Changing the way we link with SCIP

By -
As of today, I pushed support for scip 4.0.0 on both unix and windows. In the process, I have changed the way to setup or-tools to include scip at compilation time. on Unix:   untar the scipoptsuite> archive somewhere   cd <path to scipoptsuite-4.0.0>   Mac OS X:  make GMP=false READLINE=false TPI=tny   Linux:  make GMP=false READLINE=false TPI=tny USRCFLAGS=-fPIC USRCXXFLAGS=-fPIC USRCPPFLAGS=-fPIC   in or-tools, edit Makefile.local   add:     UNIX_SCIP_DIR=<path to sciptopsuite-4.0.0>/scip-4.0.0   make clean   make cc   to test it, run:      bin/integer_programming   it should display the scip solver run. on windows:   download the compiled library:     http://scip.zib.de/download.php?fname=scip-4.0.0.win.x86_64.vc14.opt.spx2.lib.zip   download the headers:      http://scip.zib.de/download/release/scipoptheaders-4.0.0.tgz   uncompress, untar everything.   create a directory that will contains what is needed   md scip-4.0.0  
Read more

Heads up, upcoming incompatibility in the CP/Routing library

By -
Hello all, I will soon push a large incompatible change in the CP and Routing libraries. The change will replace parameters (CP solver parameters, routing model parameters, routing search parameters) from struct to protobuf. This will also hide the C++ command line flags (cp_trace_search, routing_first_solution, ...) and make them redundant for controlling the behavior of the libraries. They are currently kept in the C++ library, but will be dropped out eventually. With this move, all functionalities offered by the C++ command line flags will be available from all languages (provided you implement the command line parsing part). Let's see an example (tsp) Before (in C++):    // Setting first solution heuristic (cheapest addition).    FLAGS_routing_first_solution = "PathCheapestArc";    const Assignment* solution = routing.Solve() ; After (in C++, using the command line flags compatibility) :    RoutingSearchParameters parameters =  BuildSearchParametersFro
Read more