/*
 * \file Elements.c
 * \brief: implementation of the Elements class, derived from DataSet class
 */

/*Headers: {{{*/
#ifdef HAVE_CONFIG_H
	#include <config.h>
#else
#error "Cannot compile with HAVE_CONFIG_H symbol! run configure first!"
#endif

#include <vector>
#include <functional>
#include <algorithm>
#include <iostream>

#include "./DataSet.h"
#include "../shared/shared.h"
#include "../include/include.h"
#include "../EnumDefinitions/EnumDefinitions.h"

using namespace std;
/*}}}*/

/*Object constructors and destructor*/
/*FUNCTION Elements::Elements(){{{*/
Elements::Elements(){
	enum_type=MeshElementsEnum;
	return;
}
/*}}}*/
/*FUNCTION Elements::~Elements(){{{*/
Elements::~Elements(){
	return;
}
/*}}}*/

/*Object management*/
/*FUNCTION Elements::Configure{{{*/
void Elements::Configure(Elements* elements,Loads* loads, Nodes* nodes, Vertices* vertices, Materials* materials,Parameters* parameters){

	vector<Object*>::iterator object;
	Element* element=NULL;

	for ( object=objects.begin() ; object < objects.end(); object++ ){

		element=dynamic_cast<Element*>((*object));
		element->Configure(elements,loads,nodes,materials,parameters);

	}

}
/*}}}*/
/*FUNCTION Elements::ProcessResultsUnits{{{*/
void Elements::ProcessResultsUnits(void){

	//Process results to be output in the correct units
	for(int i=0;i<this->Size();i++){
		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		element->ProcessResultsUnits();
	}
}
/*}}}*/
/*FUNCTION Elements::DeleteResults{{{*/
void Elements::DeleteResults(void){

	for (int i=0;i<this->Size();i++){
		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		element->DeleteResults();
	}
}
/*}}}*/
/*FUNCTION Elements::ResultsToPatch{{{*/
Patch* Elements::ResultsToPatch(void){ 

	/*output: */
	Patch* patch=NULL;

	/*intermediary: */
	int i;
	int count;
	int numrows;
	int numvertices;
	int numnodes;
	int max_numvertices;
	int max_numnodes;
	int element_numvertices;
	int element_numrows;
	int element_numnodes;

	/*We are going to extract from the results within the elements, the desired results , and create a table 
	 * of patch information, that will hold, for each element that computed the result that 
	 * we desire, the enum_type of the result, the step and time, the id of the element, the interpolation type, the vertices ids, and the values 
	 * at the nodes (could be different from the vertices). This will be used for visualization purposes. 
	 * For example, we could build the following patch table, for velocities: 
	 * 
	 1. on a Beam element, Vx, at step 1, time .5, element id 1, interpolation type P0 (constant), vertices ids 1 and 2, one constant value 4.5
	 VxEnum 1  .5  1 P0  1 2       4.5 NaN  NaN
	 2. on a Tria element, Vz, at step 2, time .8, element id 2, interpolation type P1 (linear), vertices ids 1 3 and 4, with values at 3 nodes 4.5, 3.2, 2.5
	 VzEnum 2  .8  2 P1  1 3 4     4.5 3.2  2.5
	 * ... etc ...
	 *
	 * So what do we need to build the table: the maximum number of vertices included in the table, 
	 * and the maximum number of nodal values, as well as the number of rows. Once we have that, 
	 * we ask the elements to fill their own row in the table, by looping on the elememnts. 
	 *
	 * We will use the Patch object, which will store all of the information needed, and will be able 
	 * to output itself to disk on its own. See object/Patch.h for format of this object.*/

	/*First, determine maximum number of vertices, nodes, and number of results: */
	numrows=0;
	numvertices=0;
	numnodes=0;

	for(i=0;i<this->Size();i++){

		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		element->PatchSize(&element_numrows,&element_numvertices,&element_numnodes);

		numrows+=element_numrows;
		if(element_numvertices>numvertices)numvertices=element_numvertices;
		if(element_numnodes>numnodes)numnodes=element_numnodes;
	}

	/*Synchronize across cluster, so as to not end up with different sizes for each patch on each cpu: */
	#ifdef _HAVE_MPI_
	MPI_Reduce (&numvertices,&max_numvertices,1,MPI_INT,MPI_MAX,0,IssmComm::GetComm() );
	MPI_Bcast(&max_numvertices,1,MPI_INT,0,IssmComm::GetComm());
	numvertices=max_numvertices;

	MPI_Reduce (&numnodes,&max_numnodes,1,MPI_INT,MPI_MAX,0,IssmComm::GetComm() );
	MPI_Bcast(&max_numnodes,1,MPI_INT,0,IssmComm::GetComm());
	numnodes=max_numnodes;
	#endif

	/*Ok, initialize Patch object: */
	patch=new Patch(numrows,numvertices,numnodes);

	/*Now, go through elements, and fill the Patch object: */
	count=0;
	for(i=0;i<this->Size();i++){
		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		element->PatchFill(&count,patch);
	}

	return patch;
}
/*}}}*/
/*FUNCTION Elements::SetCurrentConfiguration{{{*/
void Elements::SetCurrentConfiguration(Elements* elements,Loads* loads, Nodes* nodes, Vertices* vertices, Materials* materials,Parameters* parameters){

	vector<Object*>::iterator object;
	Element* element=NULL;

	for ( object=objects.begin() ; object < objects.end(); object++ ){

		element=dynamic_cast<Element*>((*object));
		element->SetCurrentConfiguration(elements,loads,nodes,materials,parameters);

	}

}
/*}}}*/
/*FUNCTION Elements::ToResults{{{*/
void Elements::ToResults(Results* results,Parameters* parameters){

	int my_rank;
	int num_procs;

	Patch               *patch              = NULL;
	int                 *resultsenums       = NULL;
	int                 *resultssizes       = NULL;
	int                 *resultssteps       = NULL;
	IssmDouble          *resultstimes       = NULL;
	IssmDouble          *vector_serial      = NULL;
	Vector<IssmDouble> *vector = NULL;
	bool                io_gather;
	bool                results_as_patches;
	int                 numberofvertices     ,numberofelements;
	int                 numberofresults      ,vectorsize;
	int                 rank;
	int                 minrank;

	/*recover my_rank:*/
	my_rank=IssmComm::GetRank();
	num_procs=IssmComm::GetSize();

	/*Recover parameters: */
	parameters->FindParam(&io_gather,SettingsIoGatherEnum);
	parameters->FindParam(&results_as_patches,SettingsResultsAsPatchesEnum);
	parameters->FindParam(&numberofvertices,MeshNumberofverticesEnum);
	parameters->FindParam(&numberofelements,MeshNumberofelementsEnum);

	if(!results_as_patches){
		/*No patch here, we prepare vectors*/

		/*Get rank of first cpu that has results*/
		#ifdef _HAVE_MPI_
		if(this->Size()) rank=my_rank;
		else rank=num_procs;
		MPI_Allreduce (&rank,&minrank,1,MPI_INT,MPI_MIN,IssmComm::GetComm());
		#else
		minrank=my_rank;
		#endif

		/*see what the first element of this partition has in stock (this is common to all partitions)*/
		if(my_rank==minrank){
			if(this->Size()==0) _error_("Cannot write results because there is no element??");
			Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(0));
			element->ListResultsInfo(&resultsenums,&resultssizes,&resultstimes,&resultssteps,&numberofresults);
		}
		#ifdef _HAVE_MPI_
		MPI_Bcast(&numberofresults,1,MPI_DOUBLE,minrank,IssmComm::GetComm());
		#endif

		/*Get out if there is no results. Otherwise broadcast info*/
		if(!numberofresults) return;
		#ifdef _HAVE_MPI_
		if(my_rank!=minrank){
			resultsenums=xNew<int>(numberofresults);
			resultssizes=xNew<int>(numberofresults);
			resultstimes=xNew<IssmDouble>(numberofresults);
			resultssteps=xNew<int>(numberofresults);
		}
		MPI_Bcast(resultsenums,numberofresults,MPI_INT,minrank,IssmComm::GetComm());
		MPI_Bcast(resultssizes,numberofresults,MPI_INT,minrank,IssmComm::GetComm());
		MPI_Bcast(resultstimes,numberofresults,MPI_DOUBLE,minrank,IssmComm::GetComm());
		MPI_Bcast(resultssteps,numberofresults,MPI_INT,minrank,IssmComm::GetComm());
		#endif

		/*Loop over all results and get nodal vector*/
		for(int i=0;i<numberofresults;i++){

			/*Get vector for result number i*/
			if(resultssizes[i]==P1Enum)      vectorsize=numberofvertices;
			else if(resultssizes[i]==P0Enum) vectorsize=numberofelements;
			else _error_("Unkown result size: " << EnumToStringx(resultssizes[i]));
			vector=new Vector<IssmDouble>(vectorsize);

			for(int j=0;j<this->Size();j++){
				Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(j));
				element->GetVectorFromResults(vector,i,resultsenums[i],resultssizes[i]);
			}
			vector->Assemble();

			/*Serialize and add to results*/
			vector_serial=vector->ToMPISerial();
			if(my_rank==0){
				/*No need to add this vector for all cpus*/
				#ifdef _HAVE_ADOLC_
				IssmPDouble* vector_serial_passive=xNew<IssmPDouble>(vectorsize);
				for(int k=0;k<vectorsize;k++)vector_serial_passive[k]=reCast<IssmPDouble>(vector_serial[k]);
				results->AddObject(new GenericExternalResult<double*>(results->Size()+1,resultsenums[i],vector_serial_passive,vectorsize,1,resultssteps[i],resultstimes[i]));
				xDelete<IssmPDouble>(vector_serial_passive);
				#else
				results->AddObject(new GenericExternalResult<double*>(results->Size()+1,resultsenums[i],vector_serial,vectorsize,1,resultssteps[i],resultstimes[i]));
				#endif
			}

			/*clean up*/
			xdelete(&vector);
			xDelete<IssmDouble>(vector_serial);
		}
	}
	else{
		/*create patch object out of all results in this dataset: */
		patch=this->ResultsToPatch();

		/*Gather onto master cpu 0, if needed: */
		if(io_gather)patch->Gather();

		/*create result object and add to results dataset:*/
		if (patch->maxvertices && patch->maxnodes){
			results->AddObject(new GenericExternalResult<int>(results->Size()+1,PatchVerticesEnum,patch->maxvertices,1,0));
			results->AddObject(new GenericExternalResult<int>(results->Size()+1,PatchNodesEnum,   patch->maxnodes,1,0));
			#ifdef _HAVE_ADOLC_
			IssmPDouble* values_passive=xNew<IssmPDouble>(patch->numrows*patch->numcols);
			for(int k=0;k<(patch->numrows*patch->numcols);k++)values_passive[k]=reCast<IssmPDouble>(patch->values[k]);
			results->AddObject(new GenericExternalResult<double*>(results->Size()+1,PatchEnum,  values_passive,patch->numrows,patch->numcols,1,0));
			xDelete<IssmPDouble>(values_passive);
			#else
			results->AddObject(new GenericExternalResult<double*>(results->Size()+1,PatchEnum,  patch->values,patch->numrows,patch->numcols,1,0));
			#endif
		}
	}

	/*Free ressources:*/
	xDelete<int>(resultsenums);
	xDelete<int>(resultssizes);
	xDelete<int>(resultssteps);
	xDelete<IssmDouble>(resultstimes);
	delete patch;
}
/*}}}*/
/*FUNCTION Elements::MaxNumNodes{{{*/
int Elements::MaxNumNodes(void){

	int max=0;
	int allmax;
	int numnodes=0;

	/*Now go through all elements, and get how many nodes they own, unless they are clone nodes: */
	for(int i=0;i<this->Size();i++){

		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		numnodes=element->GetNumberOfNodes();
		if(numnodes>max)max=numnodes;
	}

	/*Grab max of all cpus: */
	#ifdef _HAVE_MPI_
	MPI_Allreduce((void*)&max,(void*)&allmax,1,MPI_INT,MPI_MAX,IssmComm::GetComm());
	max=allmax;
	#endif

	return max;
}
/*}}}*/
/*FUNCTION Elements::NumberOfElements{{{*/
int Elements::NumberOfElements(void){

	int local_nelem;
	int numberofelements;

	local_nelem=this->Size();
	#ifdef _HAVE_MPI_
	MPI_Allreduce ( (void*)&local_nelem,(void*)&numberofelements,1,MPI_INT,MPI_SUM,IssmComm::GetComm());
	#else
	numberofelements=local_nelem;
	#endif

	return numberofelements;
}
/*}}}*/
/*FUNCTION Elements::InputCopy{{{*/
void Elements::InputDuplicate(int input_enum,int output_enum){

	for(int i=0;i<this->Size();i++){
		Element* element=dynamic_cast<Element*>(this->GetObjectByOffset(i));
		element->InputDuplicate(input_enum,output_enum);
	}
}
/*}}}*/