source: src/Actions/FragmentationAction/FragmentationAutomationAction.cpp@ 852ea3

/*
 * Project: MoleCuilder
 * Description: creates and alters molecular systems
 * Copyright (C)  2010-2012 University of Bonn. All rights reserved.
 * Please see the LICENSE file or "Copyright notice" in builder.cpp for details.
 */

/*
 * FragmentationAutomationAction.cpp
 *
 *  Created on: May 18, 2012
 *      Author: heber
 */

// include config.h
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#include <boost/archive/text_iarchive.hpp>
// boost asio needs specific operator new
#include <boost/asio.hpp>

#include "CodePatterns/MemDebug.hpp"

#include "CodePatterns/Info.hpp"
#include "CodePatterns/Log.hpp"
#include "JobMarket/Controller/FragmentController.hpp"
#include "JobMarket/Jobs/FragmentJob.hpp"

#include "Atom/atom.hpp"
#include "Fragmentation/EnergyMatrix.hpp"
#include "Fragmentation/ForceMatrix.hpp"
#include "Fragmentation/Fragmentation.hpp"
#include "Fragmentation/HydrogenSaturation_enum.hpp"
#include "Fragmentation/KeySetsContainer.hpp"
#include "Graph/DepthFirstSearchAnalysis.hpp"
#include "Jobs/MPQCCommandJob.hpp"
#include "molecule.hpp"
#include "World.hpp"

#include <iostream>
#include <string>
#include <vector>

#include "Actions/FragmentationAction/FragmentationAutomationAction.hpp"

using namespace MoleCuilder;

// and construct the stuff
#include "FragmentationAutomationAction.def"
#include "Action_impl_pre.hpp"
/** =========== define the function ====================== */

class controller_AddOn;

// needs to be defined for using the FragmentController
controller_AddOn *getAddOn()
{
  return NULL;
}

/** Creates a MPQCCommandJob with argument \a filename.
 *
 * @param jobs created job is added to this vector
 * @param command mpqc command to execute
 * @param filename filename being argument to job
 * @param nextid id for this job
 */
void parsejob(
    std::vector<FragmentJob::ptr> &jobs,
    const std::string &command,
    const std::string &filename,
    const JobId_t nextid)
{
  std::ifstream file;
  file.open(filename.c_str());
  ASSERT( file.good(), "parsejob() - file "+filename+" does not exist.");
  std::string output((std::istreambuf_iterator<char>(file)),
      std::istreambuf_iterator<char>());
  FragmentJob::ptr testJob( new MPQCCommandJob(output, nextid, command) );
  jobs.push_back(testJob);
  file.close();
  LOG(1, "INFO: Added MPQCCommandJob from file "+filename+".");
}

/** Helper function to get number of atoms somehow.
 *
 * Here, we just parse the number of lines in the adjacency file as
 * it should correspond to the number of atoms, except when some atoms
 * are not bonded, but then fragmentation makes no sense.
 *
 * @param path path to the adjacency file
 */
size_t getNoAtomsFromAdjacencyFile(const std::string &path)
{
  size_t NoAtoms = 0;

  // parse in special file to get atom count (from line count)
  std::string filename(path);
  filename += FRAGMENTPREFIX;
  filename += ADJACENCYFILE;
  std::ifstream adjacency(filename.c_str());
  if (adjacency.fail()) {
    LOG(0, endl << "getNoAtomsFromAdjacencyFile() - Unable to open " << filename << ", is the directory correct?");
    return false;
  }
  std::string buffer;
  while (getline(adjacency, buffer))
    NoAtoms++;
  LOG(1, "INFO: There are " << NoAtoms << " atoms.");

  return NoAtoms;
}


/** Print MPQCData from received results.
 *
 * @param results received results to extract MPQCData from
 * @param KeySetFilename filename with keysets to associate forces correctly
 * @param NoAtoms total number of atoms
 */
bool printReceivedMPQCResults(
    const std::vector<FragmentResult::ptr> &results,
    const std::string &KeySetFilename,
    size_t NoAtoms)
{
  EnergyMatrix Energy;
  EnergyMatrix EnergyFragments;
  ForceMatrix Force;
  ForceMatrix ForceFragments;
  KeySetsContainer KeySet;

  // align fragments
  std::map< JobId_t, size_t > MatrixNrLookup;
  size_t FragmentCounter = 0;
  {
    // bring ids in order ...
    typedef std::map< JobId_t, FragmentResult::ptr> IdResultMap_t;
    IdResultMap_t IdResultMap;
    for (std::vector<FragmentResult::ptr>::const_iterator iter = results.begin();
        iter != results.end(); ++iter) {
  #ifndef NDEBUG
      std::pair< IdResultMap_t::iterator, bool> inserter =
  #endif
      IdResultMap.insert( make_pair((*iter)->getId(), *iter) );
      ASSERT( inserter.second,
          "printReceivedMPQCResults() - two results have same id "
          +toString((*iter)->getId())+".");
    }
    // ... and fill lookup
    for(IdResultMap_t::const_iterator iter = IdResultMap.begin();
        iter != IdResultMap.end(); ++iter)
      MatrixNrLookup.insert( make_pair(iter->first, FragmentCounter++) );
  }
  LOG(1, "INFO: There are " << FragmentCounter << " fragments.");

  // extract results
  std::vector<MPQCData> fragmentData(results.size());
  MPQCData combinedData;

  LOG(2, "DEBUG: Parsing now through " << results.size() << " results.");
  for (std::vector<FragmentResult::ptr>::const_iterator iter = results.begin();
      iter != results.end(); ++iter) {
    LOG(1, "RESULT: job #"+toString((*iter)->getId())+": "+toString((*iter)->result));
    MPQCData extractedData;
    std::stringstream inputstream((*iter)->result);
    LOG(2, "DEBUG: First 50 characters FragmentResult's string: "+(*iter)->result.substr(0, 50));
    boost::archive::text_iarchive ia(inputstream);
    ia >> extractedData;
    LOG(1, "INFO: extracted data is " << extractedData << ".");

    // place results into EnergyMatrix ...
    {
      MatrixContainer::MatrixArray matrix;
      matrix.resize(1);
      matrix[0].resize(1, extractedData.energy);
      if (!Energy.AddMatrix(
          std::string("MPQCJob ")+toString((*iter)->getId()),
          matrix,
          MatrixNrLookup[(*iter)->getId()])) {
        ELOG(1, "Adding energy matrix failed.");
        return false;
      }
    }
    // ... and ForceMatrix (with two empty columns in front)
    {
      MatrixContainer::MatrixArray matrix;
      const size_t rows = extractedData.forces.size();
      matrix.resize(rows);
      for (size_t i=0;i<rows;++i) {
        const size_t columns = 2+extractedData.forces[i].size();
        matrix[i].resize(columns, 0.);
  //      for (size_t j=0;j<2;++j)
  //        matrix[i][j] = 0.;
        for (size_t j=2;j<columns;++j)
          matrix[i][j] = extractedData.forces[i][j-2];
      }
      if (!Force.AddMatrix(
          std::string("MPQCJob ")+toString((*iter)->getId()),
          matrix,
          MatrixNrLookup[(*iter)->getId()])) {
        ELOG(1, "Adding force matrix failed.");
        return false;
      }
    }
  }
  // add one more matrix (not required for energy)
  MatrixContainer::MatrixArray matrix;
  matrix.resize(1);
  matrix[0].resize(1, 0.);
  if (!Energy.AddMatrix(std::string("MPQCJob total"), matrix, FragmentCounter))
    return false;
  // but for energy because we need to know total number of atoms
  matrix.resize(NoAtoms);
  for (size_t i = 0; i< NoAtoms; ++i)
    matrix[i].resize(2+NDIM, 0.);
  if (!Force.AddMatrix(std::string("MPQCJob total"), matrix, FragmentCounter))
    return false;


  // combine all found data
  if (!Energy.InitialiseIndices()) return false;

  if (!Force.ParseIndices(KeySetFilename.c_str())) return false;

  if (!KeySet.ParseKeySets(KeySetFilename.c_str(), Force.RowCounter, Force.MatrixCounter)) return false;

  if (!KeySet.ParseManyBodyTerms()) return false;

  if (!EnergyFragments.AllocateMatrix(Energy.Header, Energy.MatrixCounter, Energy.RowCounter, Energy.ColumnCounter)) return false;
  if (!ForceFragments.AllocateMatrix(Force.Header, Force.MatrixCounter, Force.RowCounter, Force.ColumnCounter)) return false;

  if(!Energy.SetLastMatrix(0., 0)) return false;
  if(!Force.SetLastMatrix(0., 2)) return false;

  for (int BondOrder=0;BondOrder<KeySet.Order;BondOrder++) {
    // --------- sum up energy --------------------
    LOG(1, "INFO: Summing energy of order " << BondOrder+1 << " ...");
    if (!EnergyFragments.SumSubManyBodyTerms(Energy, KeySet, BondOrder)) return false;
    if (!Energy.SumSubEnergy(EnergyFragments, NULL, KeySet, BondOrder, 1.)) return false;

    // --------- sum up Forces --------------------
    LOG(1, "INFO: Summing forces of order " << BondOrder+1 << " ...");
    if (!ForceFragments.SumSubManyBodyTerms(Force, KeySet, BondOrder)) return false;
    if (!Force.SumSubForces(ForceFragments, KeySet, BondOrder, 1.)) return false;
  }

  // for debugging print resulting energy and forces
  LOG(1, "INFO: Resulting energy is " << Energy.Matrix[ FragmentCounter ][0][0]);
  std::stringstream output;
  for (int i=0; i< Force.RowCounter[FragmentCounter]; ++i) {
    for (int j=0; j< Force.ColumnCounter[FragmentCounter]; ++j)
      output << Force.Matrix[ FragmentCounter ][i][j] << " ";
    output << "\n";
  }
  LOG(1, "INFO: Resulting forces are " << std::endl << output.str());

  return true;
}

Action::state_ptr FragmentationFragmentationAutomationAction::performCall() {
  boost::asio::io_service io_service;
  FragmentController controller(io_service);

  // TODO: Have io_service run in second thread and merge with current again eventually

  // Phase One: obtain ids
  std::vector< boost::filesystem::path > jobfiles = params.jobfiles.get();
  controller.requestIds(params.host.get(), params.port.get(), jobfiles.size());
  {
    io_service.reset();
    Info info("io_service: Phase One");
    io_service.run();
  }
  // Phase Two: create and add jobs
  {
    std::vector<FragmentJob::ptr> jobs;
    for (std::vector< boost::filesystem::path >::const_iterator iter = jobfiles .begin();
        iter != jobfiles .end(); ++iter) {
      const std::string &filename = (*iter).string();
      if (boost::filesystem::exists(filename)) {
        const JobId_t next_id = controller.getAvailableId();
        LOG(1, "INFO: Creating MPQCCommandJob with filename'"
            +filename+"', and id "+toString(next_id)+".");
        parsejob(jobs, params.executable.get().string(), filename, next_id);
      } else {
        ELOG(1, "Fragment job "+filename+" does not exist.");
        return Action::failure;
      }
    }
    controller.addJobs(jobs);
    controller.sendJobs(params.host.get(), params.port.get());
  }
  {
    io_service.reset();
    Info info("io_service: Phase Two");
    io_service.run();
  }
  // Phase Three: calculate result
  size_t NoCalculatedResults = 0;
  while (NoCalculatedResults != jobfiles.size()) {
    // wait a bit
    boost::asio::deadline_timer timer(io_service);
    timer.expires_from_now(boost::posix_time::milliseconds(500));
    timer.wait();
    // then request status
    controller.checkResults(params.host.get(), params.port.get());
    {
      io_service.reset();
      Info info("io_service: Phase Three");
      io_service.run();
    }
    const std::pair<size_t, size_t> JobStatus = controller.getJobStatus();
    LOG(1, "INFO: #" << JobStatus.first << " are waiting in the queue and #" << JobStatus.second << " jobs are calculated so far.");
    NoCalculatedResults = JobStatus.second;
  }
  // Phase Three: get result
  controller.receiveResults(params.host.get(), params.port.get());
  {
    io_service.reset();
    Info info("io_service: Phase Four");
    io_service.run();
  }
  // Final phase: print result
  {
    LOG(1, "INFO: Parsing fragment files from " << params.path.get() << ".");
    std::vector<FragmentResult::ptr> results = controller.getReceivedResults();
    printReceivedMPQCResults(
        results,
        params.path.get(),
        getNoAtomsFromAdjacencyFile(params.path.get()));
  }
  size_t Exitflag = controller.getExitflag();

  return (Exitflag == 0) ? Action::success : Action::failure;
}

Action::state_ptr FragmentationFragmentationAutomationAction::performUndo(Action::state_ptr _state) {
  return Action::success;
}

Action::state_ptr FragmentationFragmentationAutomationAction::performRedo(Action::state_ptr _state){
  return Action::success;
}

bool FragmentationFragmentationAutomationAction::canUndo() {
  return false;
}

bool FragmentationFragmentationAutomationAction::shouldUndo() {
  return false;
}
/** =========== end of function ====================== */