/[escript]/trunk/escript/src/Data.cpp
ViewVC logotype

Diff of /trunk/escript/src/Data.cpp

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 790 by bcumming, Wed Jul 26 23:12:34 2006 UTC revision 2735 by jfenwick, Mon Nov 2 02:03:24 2009 UTC
# Line 1  Line 1 
 // $Id$  
1    
2  /*  /*******************************************************
3   ************************************************************  *
4   *          Copyright 2006 by ACcESS MNRF                   *  * Copyright (c) 2003-2009 by University of Queensland
5   *                                                          *  * Earth Systems Science Computational Center (ESSCC)
6   *              http://www.access.edu.au                    *  * http://www.uq.edu.au/esscc
7   *       Primary Business: Queensland, Australia            *  *
8   *  Licensed under the Open Software License version 3.0    *  * Primary Business: Queensland, Australia
9   *     http://www.opensource.org/licenses/osl-3.0.php       *  * Licensed under the Open Software License version 3.0
10   *                                                          *  * http://www.opensource.org/licenses/osl-3.0.php
11   ************************************************************  *
12  */  *******************************************************/
13    
14    
15  #include "Data.h"  #include "Data.h"
16    
17  #include "DataExpanded.h"  #include "DataExpanded.h"
18  #include "DataConstant.h"  #include "DataConstant.h"
19  #include "DataTagged.h"  #include "DataTagged.h"
20  #include "DataEmpty.h"  #include "DataEmpty.h"
21  #include "DataArray.h"  #include "DataLazy.h"
 #include "DataArrayView.h"  
 #include "DataProf.h"  
22  #include "FunctionSpaceFactory.h"  #include "FunctionSpaceFactory.h"
23  #include "AbstractContinuousDomain.h"  #include "AbstractContinuousDomain.h"
24  #include "UnaryFuncs.h"  #include "UnaryFuncs.h"
25    #include "FunctionSpaceException.h"
26    #include "EscriptParams.h"
27    
28    extern "C" {
29    #include "esysUtils/blocktimer.h"
30    }
31    
32  #include <fstream>  #include <fstream>
33  #include <algorithm>  #include <algorithm>
34  #include <vector>  #include <vector>
35  #include <functional>  #include <functional>
36    #include <sstream>  // so we can throw messages about ranks
37    
38  #include <boost/python/dict.hpp>  #include <boost/python/dict.hpp>
39  #include <boost/python/extract.hpp>  #include <boost/python/extract.hpp>
40  #include <boost/python/long.hpp>  #include <boost/python/long.hpp>
41    #include "WrappedArray.h"
42    
43  using namespace std;  using namespace std;
44  using namespace boost::python;  using namespace boost::python;
45  using namespace boost;  using namespace boost;
46  using namespace escript;  using namespace escript;
47    
48  #if defined DOPROF  // ensure the current object is not a DataLazy
49  //  // The idea was that we could add an optional warning whenever a resolve is forced
50  // global table of profiling data for all Data objects  // #define forceResolve() if (isLazy()) {#resolve();}
51  DataProf dataProfTable;  
52  #endif  #define AUTOLAZYON escriptParams.getAUTOLAZY()
53    #define MAKELAZYOP(X)   if (isLazy() || (AUTOLAZYON && m_data->isExpanded())) \
54      {\
55        DataLazy* c=new DataLazy(borrowDataPtr(),X);\
56        return Data(c);\
57      }
58    #define MAKELAZYOPOFF(X,Y) if (isLazy() || (AUTOLAZYON && m_data->isExpanded())) \
59      {\
60        DataLazy* c=new DataLazy(borrowDataPtr(),X,Y);\
61        return Data(c);\
62      }
63    
64    #define MAKELAZYOP2(X,Y,Z) if (isLazy() || (AUTOLAZYON && m_data->isExpanded())) \
65      {\
66        DataLazy* c=new DataLazy(borrowDataPtr(),X,Y,Z);\
67        return Data(c);\
68      }
69    
70    #define MAKELAZYBINSELF(R,X)   if (isLazy() || R.isLazy() || (AUTOLAZYON && (isExpanded() || R.isExpanded()))) \
71      {\
72        DataLazy* c=new DataLazy(m_data,R.borrowDataPtr(),X);\
73    /*         m_data=c->getPtr();*/     set_m_data(c->getPtr());\
74        return (*this);\
75      }
76    
77    // like the above but returns a new data rather than *this
78    #define MAKELAZYBIN(R,X)   if (isLazy() || R.isLazy() || (AUTOLAZYON && (isExpanded() || R.isExpanded()))) \
79      {\
80        DataLazy* c=new DataLazy(m_data,R.borrowDataPtr(),X);\
81        return Data(c);\
82      }
83    
84    #define MAKELAZYBIN2(L,R,X)   if (L.isLazy() || R.isLazy() || (AUTOLAZYON && (L.isExpanded() || R.isExpanded()))) \
85      {\
86        DataLazy* c=new DataLazy(L.borrowDataPtr(),R.borrowDataPtr(),X);\
87        return Data(c);\
88      }
89    
90    #define CHECK_DO_CRES escriptParams.getRESOLVE_COLLECTIVE()
91    
92    namespace
93    {
94    
95    template <class ARR>
96    inline
97    boost::python::tuple
98    pointToTuple1(const DataTypes::ShapeType& shape, ARR v, unsigned long offset)
99    {
100        using namespace boost::python;
101        using boost::python::tuple;
102        using boost::python::list;
103    
104        list l;
105        unsigned int dim=shape[0];
106        for (size_t i=0;i<dim;++i)
107        {
108        l.append(v[i+offset]);
109        }
110        return tuple(l);
111    }
112    
113    template <class ARR>
114    inline
115    boost::python::tuple
116    pointToTuple2(const DataTypes::ShapeType& shape, ARR v, unsigned long offset)
117    {
118        using namespace boost::python;
119        using boost::python::tuple;
120        using boost::python::list;
121    
122        unsigned int shape0=shape[0];
123        unsigned int shape1=shape[1];
124        list lj;
125        for (size_t j=0;j<shape0;++j)
126        {
127            list li;
128        for (size_t i=0;i<shape1;++i)
129        {
130            li.append(v[offset+DataTypes::getRelIndex(shape,j,i)]);
131        }
132        lj.append(tuple(li));
133        }
134        return tuple(lj);
135    }
136    
137    template <class ARR>
138    inline
139    boost::python::tuple
140    pointToTuple3(const DataTypes::ShapeType& shape, ARR v, unsigned long offset)
141    {
142        using namespace boost::python;
143        using boost::python::tuple;
144        using boost::python::list;
145    
146        unsigned int shape0=shape[0];
147        unsigned int shape1=shape[1];
148        unsigned int shape2=shape[2];
149    
150        list lk;
151        for (size_t k=0;k<shape0;++k)
152        {
153            list lj;
154        for (size_t j=0;j<shape1;++j)
155        {
156            list li;
157            for (size_t i=0;i<shape2;++i)
158            {
159                    li.append(v[offset+DataTypes::getRelIndex(shape,k,j,i)]);
160                }
161            lj.append(tuple(li));
162            }
163            lk.append(tuple(lj));
164        }
165        return tuple(lk);
166    }
167    
168    template <class ARR>
169    inline
170    boost::python::tuple
171    pointToTuple4(const DataTypes::ShapeType& shape, ARR v, unsigned long offset)
172    {
173        using namespace boost::python;
174        using boost::python::tuple;
175        using boost::python::list;
176    
177        unsigned int shape0=shape[0];
178        unsigned int shape1=shape[1];
179        unsigned int shape2=shape[2];
180        unsigned int shape3=shape[3];
181    
182        list ll;
183        for (size_t l=0;l<shape0;++l)
184        {
185            list lk;
186        for (size_t k=0;k<shape1;++k)
187        {
188                list lj;
189                for (size_t j=0;j<shape2;++j)
190                {
191                    list li;
192                    for (size_t i=0;i<shape3;++i)
193                    {
194                        li.append(v[offset+DataTypes::getRelIndex(shape,l,k,j,i)]);
195                    }
196                    lj.append(tuple(li));
197                }
198                lk.append(tuple(lj));
199        }
200            ll.append(tuple(lk));
201        }
202        return tuple(ll);
203    }
204    
205    
206    // This should be safer once the DataC RO changes have been brought in
207    template <class ARR>
208    boost::python::tuple
209    pointToTuple( const DataTypes::ShapeType& shape,ARR v)
210    {
211       using namespace boost::python;
212       using boost::python::list;
213       int rank=shape.size();
214       if (rank==0)
215       {
216        return make_tuple(v[0]);
217       }
218       else if (rank==1)
219       {
220            return pointToTuple1(shape,v,0);
221       }
222       else if (rank==2)
223       {
224        return pointToTuple2(shape,v,0);
225       }
226       else if (rank==3)
227       {
228        return pointToTuple3(shape,v,0);
229       }
230       else if (rank==4)
231       {
232        return pointToTuple4(shape,v,0);
233       }
234       else
235         throw DataException("Unknown rank in pointToTuple.");
236    }
237    
238    }  // anonymous namespace
239    
240  Data::Data()  Data::Data()
241        : m_shared(false), m_lazy(false)
242  {  {
243    //    //
244    // Default data is type DataEmpty    // Default data is type DataEmpty
245    DataAbstract* temp=new DataEmpty();    DataAbstract* temp=new DataEmpty();
246    shared_ptr<DataAbstract> temp_data(temp);  //   m_data=temp->getPtr();
247    m_data=temp_data;    set_m_data(temp->getPtr());
248    m_protected=false;    m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
249  }  }
250    
251  Data::Data(double value,  Data::Data(double value,
252             const tuple& shape,             const tuple& shape,
253             const FunctionSpace& what,             const FunctionSpace& what,
254             bool expanded)             bool expanded)
255        : m_shared(false), m_lazy(false)
256  {  {
257    DataArrayView::ShapeType dataPointShape;    DataTypes::ShapeType dataPointShape;
258    for (int i = 0; i < shape.attr("__len__")(); ++i) {    for (int i = 0; i < shape.attr("__len__")(); ++i) {
259      dataPointShape.push_back(extract<const int>(shape[i]));      dataPointShape.push_back(extract<const int>(shape[i]));
260    }    }
261    DataArray temp(dataPointShape,value);  
262    initialise(temp.getView(),what,expanded);    int len = DataTypes::noValues(dataPointShape);
263      DataVector temp_data(len,value,len);
264      initialise(temp_data, dataPointShape, what, expanded);
265    m_protected=false;    m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
266  }  }
267    
268  Data::Data(double value,  Data::Data(double value,
269         const DataArrayView::ShapeType& dataPointShape,         const DataTypes::ShapeType& dataPointShape,
270         const FunctionSpace& what,         const FunctionSpace& what,
271             bool expanded)             bool expanded)
272        : m_shared(false), m_lazy(false)
273  {  {
274    DataArray temp(dataPointShape,value);    int len = DataTypes::noValues(dataPointShape);
275    pair<int,int> dataShape=what.getDataShape();    DataVector temp_data(len,value,len);
276    initialise(temp.getView(),what,expanded);    initialise(temp_data, dataPointShape, what, expanded);
277    m_protected=false;    m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
278  }  }
279    
280  Data::Data(const Data& inData)  Data::Data(const Data& inData)
281        : m_shared(false), m_lazy(false)
282  {  {
283    m_data=inData.m_data;    set_m_data(inData.m_data);
284    m_protected=inData.isProtected();    m_protected=inData.isProtected();
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
285  }  }
286    
287    
288  Data::Data(const Data& inData,  Data::Data(const Data& inData,
289             const DataArrayView::RegionType& region)             const DataTypes::RegionType& region)
290        : m_shared(false), m_lazy(false)
291  {  {
292      DataAbstract_ptr dat=inData.m_data;
293      if (inData.isLazy())
294      {
295        dat=inData.m_data->resolve();
296      }
297      else
298      {
299        dat=inData.m_data;
300      }
301    //    //
302    // Create Data which is a slice of another Data    // Create Data which is a slice of another Data
303    DataAbstract* tmp = inData.m_data->getSlice(region);    DataAbstract* tmp = dat->getSlice(region);
304    shared_ptr<DataAbstract> temp_data(tmp);    set_m_data(DataAbstract_ptr(tmp));
   m_data=temp_data;  
305    m_protected=false;    m_protected=false;
306  #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
307  }  }
308    
309  Data::Data(const Data& inData,  Data::Data(const Data& inData,
310             const FunctionSpace& functionspace)             const FunctionSpace& functionspace)
311        : m_shared(false), m_lazy(false)
312  {  {
313  #if defined DOPROF    if (inData.isEmpty())
314    // create entry in global profiling table for this object    {
315    profData = dataProfTable.newData();      throw DataException("Error - will not interpolate for instances of DataEmpty.");
316  #endif    }
317    if (inData.getFunctionSpace()==functionspace) {    if (inData.getFunctionSpace()==functionspace) {
318      m_data=inData.m_data;      set_m_data(inData.m_data);
319    } else {    }
320      #if defined DOPROF    else
321      profData->interpolate++;    {
322      #endif  
323      Data tmp(0,inData.getPointDataView().getShape(),functionspace,true);      if (inData.isConstant()) {  // for a constant function, we just need to use the new function space
324      // Note: Must use a reference or pointer to a derived object        if (!inData.probeInterpolation(functionspace))
325      // in order to get polymorphic behaviour. Shouldn't really        {           // Even though this is constant, we still need to check whether interpolation is allowed
326      // be able to create an instance of AbstractDomain but that was done      throw FunctionSpaceException("Cannot interpolate across to the domain of the specified FunctionSpace. (DataConstant)");
327      // as a boost:python work around which may no longer be required.        }
328      const AbstractDomain& inDataDomain=inData.getDomain();        // if the data is not lazy, this will just be a cast to DataReady
329      if  (inDataDomain==functionspace.getDomain()) {        DataReady_ptr dr=inData.m_data->resolve();
330        inDataDomain.interpolateOnDomain(tmp,inData);        DataConstant* dc=new DataConstant(functionspace,inData.m_data->getShape(),dr->getVectorRO());
331    //       m_data=DataAbstract_ptr(dc);
332          set_m_data(DataAbstract_ptr(dc));
333      } else {      } else {
334        inDataDomain.interpolateACross(tmp,inData);        Data tmp(0,inData.getDataPointShape(),functionspace,true);
335          // Note: Must use a reference or pointer to a derived object
336          // in order to get polymorphic behaviour. Shouldn't really
337          // be able to create an instance of AbstractDomain but that was done
338          // as a boost:python work around which may no longer be required.
339          /*const AbstractDomain& inDataDomain=inData.getDomain();*/
340          const_Domain_ptr inDataDomain=inData.getDomain();
341          if  (inDataDomain==functionspace.getDomain()) {
342            inDataDomain->interpolateOnDomain(tmp,inData);
343          } else {
344            inDataDomain->interpolateACross(tmp,inData);
345          }
346    //       m_data=tmp.m_data;
347          set_m_data(tmp.m_data);
348      }      }
     m_data=tmp.m_data;  
349    }    }
350    m_protected=false;    m_protected=false;
351  }  }
352    
353  Data::Data(const DataTagged::TagListType& tagKeys,  Data::Data(DataAbstract* underlyingdata)
354             const DataTagged::ValueListType & values,      : m_shared(false), m_lazy(false)
            const DataArrayView& defaultValue,  
            const FunctionSpace& what,  
            bool expanded)  
355  {  {
356    DataAbstract* temp=new DataTagged(tagKeys,values,defaultValue,what);      set_m_data(underlyingdata->getPtr());
357    shared_ptr<DataAbstract> temp_data(temp);      m_protected=false;
   m_data=temp_data;  
   m_protected=false;  
   if (expanded) {  
     expand();  
   }  
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
358  }  }
359    
360  Data::Data(const numeric::array& value,  Data::Data(DataAbstract_ptr underlyingdata)
361         const FunctionSpace& what,      : m_shared(false), m_lazy(false)
            bool expanded)  
362  {  {
363    initialise(value,what,expanded);      set_m_data(underlyingdata);
364    m_protected=false;      m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
365  }  }
366    
367  Data::Data(const DataArrayView& value,  Data::Data(const DataTypes::ValueType& value,
368         const FunctionSpace& what,           const DataTypes::ShapeType& shape,
369             bool expanded)                   const FunctionSpace& what,
370                     bool expanded)
371        : m_shared(false), m_lazy(false)
372  {  {
373    initialise(value,what,expanded);     initialise(value,shape,what,expanded);
374    m_protected=false;     m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
375  }  }
376    
377    
378  Data::Data(const object& value,  Data::Data(const object& value,
379         const FunctionSpace& what,         const FunctionSpace& what,
380             bool expanded)             bool expanded)
381        : m_shared(false), m_lazy(false)
382  {  {
383    numeric::array asNumArray(value);    WrappedArray w(value);
384    initialise(asNumArray,what,expanded);    initialise(w,what,expanded);
385    m_protected=false;    m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
386  }  }
387    
388    
389  Data::Data(const object& value,  Data::Data(const object& value,
390             const Data& other)             const Data& other)
391        : m_shared(false), m_lazy(false)
392  {  {
393    //    WrappedArray w(value);
394    // Create DataConstant using the given value and all other parameters  
395    // copied from other. If value is a rank 0 object this Data    // extract the shape of the array
396    // will assume the point data shape of other.    const DataTypes::ShapeType& tempShape=w.getShape();
397    DataArray temp(value);    if (w.getRank()==0) {
398    if (temp.getView().getRank()==0) {  
399      //  
400      // Create a DataArray with the scalar value for all elements      // get the space for the data vector
401      DataArray temp2(other.getPointDataView().getShape(),temp.getView()());      int len1 = DataTypes::noValues(tempShape);
402      initialise(temp2.getView(),other.getFunctionSpace(),false);      DataVector temp_data(len1, 0.0, len1);
403        temp_data.copyFromArray(w,1);
404    
405        int len = DataTypes::noValues(other.getDataPointShape());
406    
407        DataVector temp2_data(len, temp_data[0], len);
408        DataConstant* t=new DataConstant(other.getFunctionSpace(),other.getDataPointShape(),temp2_data);
409    //     m_data=DataAbstract_ptr(t);
410        set_m_data(DataAbstract_ptr(t));
411    
412    } else {    } else {
413      //      //
414      // Create a DataConstant with the same sample shape as other      // Create a DataConstant with the same sample shape as other
415      initialise(temp.getView(),other.getFunctionSpace(),false);      DataConstant* t=new DataConstant(w,other.getFunctionSpace());
416    //     m_data=DataAbstract_ptr(t);
417        set_m_data(DataAbstract_ptr(t));
418    }    }
419    m_protected=false;    m_protected=false;
 #if defined DOPROF  
   // create entry in global profiling table for this object  
   profData = dataProfTable.newData();  
 #endif  
420  }  }
421    
422  Data::~Data()  Data::~Data()
423  {  {
424      set_m_data(DataAbstract_ptr());
425    }
426    
427    
428    // only call in thread safe contexts.
429    // This method should be atomic
430    void Data::set_m_data(DataAbstract_ptr p)
431    {
432      if (m_data.get()!=0)  // release old ownership
433      {
434        m_data->removeOwner(this);
435      }
436      if (p.get()!=0)
437      {
438        m_data=p;
439        m_data->addOwner(this);
440        m_shared=m_data->isShared();
441        m_lazy=m_data->isLazy();
442      }
443    }
444    
445    void Data::initialise(const WrappedArray& value,
446                     const FunctionSpace& what,
447                     bool expanded)
448    {
449      //
450      // Construct a Data object of the appropriate type.
451      // Construct the object first as there seems to be a bug which causes
452      // undefined behaviour if an exception is thrown during construction
453      // within the shared_ptr constructor.
454      if (expanded) {
455        DataAbstract* temp=new DataExpanded(value, what);
456    //     m_data=temp->getPtr();
457        set_m_data(temp->getPtr());
458      } else {
459        DataAbstract* temp=new DataConstant(value, what);
460    //     m_data=temp->getPtr();
461        set_m_data(temp->getPtr());
462      }
463    }
464    
465    
466    void
467    Data::initialise(const DataTypes::ValueType& value,
468             const DataTypes::ShapeType& shape,
469                     const FunctionSpace& what,
470                     bool expanded)
471    {
472      //
473      // Construct a Data object of the appropriate type.
474      // Construct the object first as there seems to be a bug which causes
475      // undefined behaviour if an exception is thrown during construction
476      // within the shared_ptr constructor.
477      if (expanded) {
478        DataAbstract* temp=new DataExpanded(what, shape, value);
479    //     m_data=temp->getPtr();
480        set_m_data(temp->getPtr());
481      } else {
482        DataAbstract* temp=new DataConstant(what, shape, value);
483    //     m_data=temp->getPtr();
484        set_m_data(temp->getPtr());
485      }
486  }  }
487    
488    
489  escriptDataC  escriptDataC
490  Data::getDataC()  Data::getDataC()
491  {  {
# Line 247  Data::getDataC() const Line 502  Data::getDataC() const
502    return temp;    return temp;
503  }  }
504    
505    size_t
506    Data::getSampleBufferSize() const
507    {
508      return m_data->getSampleBufferSize();
509    }
510    
511    
512  const boost::python::tuple  const boost::python::tuple
513  Data::getShapeTuple() const  Data::getShapeTuple() const
514  {  {
515    const DataArrayView::ShapeType& shape=getDataPointShape();    const DataTypes::ShapeType& shape=getDataPointShape();
516    switch(getDataPointRank()) {    switch(getDataPointRank()) {
517       case 0:       case 0:
518          return make_tuple();          return make_tuple();
# Line 267  Data::getShapeTuple() const Line 529  Data::getShapeTuple() const
529    }    }
530  }  }
531    
532    
533    // The different name is needed because boost has trouble with overloaded functions.
534    // It can't work out what type the function is based soley on its name.
535    // There are ways to fix this involving creating function pointer variables for each form
536    // but there doesn't seem to be a need given that the methods have the same name from the python point of view
537    Data
538    Data::copySelf()
539    {
540       DataAbstract* temp=m_data->deepCopy();
541       return Data(temp);
542    }
543    
544  void  void
545  Data::copy(const Data& other)  Data::copy(const Data& other)
546  {  {
547    //    DataAbstract* temp=other.m_data->deepCopy();
548    // Perform a deep copy    DataAbstract_ptr p=temp->getPtr();
549    //   m_data=p;
550      set_m_data(p);
551    }
552    
553    
554    Data
555    Data::delay()
556    {
557      if (!isLazy())
558    {    {
559      DataExpanded* temp=dynamic_cast<DataExpanded*>(other.m_data.get());        DataLazy* dl=new DataLazy(m_data);
560      if (temp!=0) {        return Data(dl);
       //  
       // Construct a DataExpanded copy  
       DataAbstract* newData=new DataExpanded(*temp);  
       shared_ptr<DataAbstract> temp_data(newData);  
       m_data=temp_data;  
       return;  
     }  
561    }    }
562      return *this;
563    }
564    
565    void
566    Data::delaySelf()
567    {
568      if (!isLazy())
569    {    {
570      DataTagged* temp=dynamic_cast<DataTagged*>(other.m_data.get());  //  m_data=(new DataLazy(m_data))->getPtr();
571      if (temp!=0) {      set_m_data((new DataLazy(m_data))->getPtr());
572        //    }
573        // Construct a DataTagged copy  }
574        DataAbstract* newData=new DataTagged(*temp);  
575        shared_ptr<DataAbstract> temp_data(newData);  
576        m_data=temp_data;  // For lazy data, it would seem that DataTagged will need to be treated differently since even after setting all tags
577        return;  // to zero, all the tags from all the DataTags would be in the result.
578      }  // However since they all have the same value (0) whether they are there or not should not matter.
579    // So I have decided that for all types this method will create a constant 0.
580    // It can be promoted up as required.
581    // A possible efficiency concern might be expanded->constant->expanded which has an extra memory management
582    // but we can deal with that if it arises.
583    //
584    void
585    Data::setToZero()
586    {
587      if (isEmpty())
588      {
589         throw DataException("Error - Operations not permitted on instances of DataEmpty.");
590    }    }
591      if (isLazy())
592    {    {
593      DataConstant* temp=dynamic_cast<DataConstant*>(other.m_data.get());       DataTypes::ValueType v(getNoValues(),0);
594      if (temp!=0) {       DataConstant* dc=new DataConstant(getFunctionSpace(),getDataPointShape(),v);
595        //       DataLazy* dl=new DataLazy(dc->getPtr());
596        // Construct a DataConstant copy       set_m_data(dl->getPtr());
       DataAbstract* newData=new DataConstant(*temp);  
       shared_ptr<DataAbstract> temp_data(newData);  
       m_data=temp_data;  
       return;  
     }  
597    }    }
598      else
599    {    {
600      DataEmpty* temp=dynamic_cast<DataEmpty*>(other.m_data.get());       exclusiveWrite();
601      if (temp!=0) {       m_data->setToZero();
       //  
       // Construct a DataEmpty copy  
       DataAbstract* newData=new DataEmpty();  
       shared_ptr<DataAbstract> temp_data(newData);  
       m_data=temp_data;  
       return;  
     }  
602    }    }
   throw DataException("Error - Copy not implemented for this Data type.");  
603  }  }
604    
605    
606  void  void
607  Data::copyWithMask(const Data& other,  Data::copyWithMask(const Data& other,
608                     const Data& mask)                     const Data& mask)
609  {  {
610    Data mask1;    // 1. Interpolate if required so all Datas use the same FS as this
611    Data mask2;    // 2. Tag or Expand so that all Data's are the same type
612      // 3. Iterate over the data vectors copying values where mask is >0
613      if (other.isEmpty() || mask.isEmpty())
614      {
615        throw DataException("Error - copyWithMask not permitted using instances of DataEmpty.");
616      }
617      Data other2(other);
618      Data mask2(mask);
619      other2.resolve();
620      mask2.resolve();
621      this->resolve();
622      FunctionSpace myFS=getFunctionSpace();
623      FunctionSpace oFS=other2.getFunctionSpace();
624      FunctionSpace mFS=mask2.getFunctionSpace();
625      if (oFS!=myFS)
626      {
627         if (other2.probeInterpolation(myFS))
628         {
629        other2=other2.interpolate(myFS);
630         }
631         else
632         {
633        throw DataException("Error - copyWithMask: other FunctionSpace is not compatible with this one.");
634         }
635      }
636      if (mFS!=myFS)
637      {
638         if (mask2.probeInterpolation(myFS))
639         {
640        mask2=mask2.interpolate(myFS);
641         }
642         else
643         {
644        throw DataException("Error - copyWithMask: mask FunctionSpace is not compatible with this one.");
645         }
646      }
647                // Ensure that all args have the same type
648      if (this->isExpanded() || mask2.isExpanded() || other2.isExpanded())
649      {
650        this->expand();
651        other2.expand();
652        mask2.expand();
653      }
654      else if (this->isTagged() || mask2.isTagged() || other2.isTagged())
655      {
656        this->tag();
657        other2.tag();
658        mask2.tag();
659      }
660      else if (this->isConstant() && mask2.isConstant() && other2.isConstant())
661      {
662      }
663      else
664      {
665        throw DataException("Error - Unknown DataAbstract passed to copyWithMask.");
666      }
667      unsigned int selfrank=getDataPointRank();
668      unsigned int otherrank=other2.getDataPointRank();
669      unsigned int maskrank=mask2.getDataPointRank();
670      if ((selfrank==0) && (otherrank>0 || maskrank>0))
671      {
672        // to get here we must be copying from a large object into a scalar
673        // I am not allowing this.
674        // If you are calling copyWithMask then you are considering keeping some existing values
675        // and so I'm going to assume that you don't want your data objects getting a new shape.
676        throw DataException("Attempt to copyWithMask into a scalar from an object or mask with rank>0.");
677      }
678      exclusiveWrite();
679      // Now we iterate over the elements
680      DataVector& self=getReady()->getVectorRW();;
681      const DataVector& ovec=other2.getReadyPtr()->getVectorRO();
682      const DataVector& mvec=mask2.getReadyPtr()->getVectorRO();
683    
684    mask1 = mask.wherePositive();    if ((selfrank>0) && (otherrank==0) &&(maskrank==0))
685    mask2.copy(mask1);    {
686        // Not allowing this combination.
687        // it is not clear what the rank of the target should be.
688        // Should it be filled with the scalar (rank stays the same);
689        // or should the target object be reshaped to be a scalar as well.
690        throw DataException("Attempt to copyWithMask from scalar mask and data into non-scalar target.");
691      }
692      if ((selfrank>0) && (otherrank>0) &&(maskrank==0))
693      {
694        if (mvec[0]>0)      // copy whole object if scalar is >0
695        {
696            copy(other);
697        }
698        return;
699      }
700      if (isTagged())       // so all objects involved will also be tagged
701      {
702        // note the !
703        if (!((getDataPointShape()==mask2.getDataPointShape()) &&
704            ((other2.getDataPointShape()==mask2.getDataPointShape()) || (otherrank==0))))
705        {
706            throw DataException("copyWithMask, shape mismatch.");
707        }
708    
709        // We need to consider the possibility that tags are missing or in the wrong order
710        // My guiding assumption here is: All tagged Datas are assumed to have the default value for
711        // all tags which are not explicitly defined
712    
713        const DataTagged* mptr=dynamic_cast<const DataTagged*>(mask2.m_data.get());
714        const DataTagged* optr=dynamic_cast<const DataTagged*>(other2.m_data.get());
715        DataTagged* tptr=dynamic_cast<DataTagged*>(m_data.get());
716    
717        // first, add any tags missing from other or mask
718        const DataTagged::DataMapType& olookup=optr->getTagLookup();
719            const DataTagged::DataMapType& mlookup=mptr->getTagLookup();
720        const DataTagged::DataMapType& tlookup=tptr->getTagLookup();
721        DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
722        for (i=olookup.begin();i!=olookup.end();i++)
723        {
724               tptr->addTag(i->first);
725            }
726            for (i=mlookup.begin();i!=mlookup.end();i++) {
727               tptr->addTag(i->first);
728            }
729        // now we know that *this has all the required tags but they aren't guaranteed to be in
730        // the same order
731    
732    mask1 *= other;      // There are two possibilities: 1. all objects have the same rank. 2. other is a scalar
733    mask2 *= *this;      if ((selfrank==otherrank) && (otherrank==maskrank))
734    mask2 = *this - mask2;      {
735            for (i=tlookup.begin();i!=tlookup.end();i++)
736            {
737                // get the target offset
738                DataTypes::ValueType::size_type toff=tptr->getOffsetForTag(i->first);
739                    DataTypes::ValueType::size_type moff=mptr->getOffsetForTag(i->first);
740                DataTypes::ValueType::size_type ooff=optr->getOffsetForTag(i->first);
741                for (int j=0;j<getDataPointSize();++j)
742                {
743                    if (mvec[j+moff]>0)
744                    {
745                        self[j+toff]=ovec[j+ooff];
746                    }
747                }
748                }
749            // now for the default value
750            for (int j=0;j<getDataPointSize();++j)
751            {
752                if (mvec[j+mptr->getDefaultOffset()]>0)
753                {
754                    self[j+tptr->getDefaultOffset()]=ovec[j+optr->getDefaultOffset()];
755                }
756            }
757        }
758        else    // other is a scalar
759        {
760            for (i=tlookup.begin();i!=tlookup.end();i++)
761            {
762                // get the target offset
763                DataTypes::ValueType::size_type toff=tptr->getOffsetForTag(i->first);
764                    DataTypes::ValueType::size_type moff=mptr->getOffsetForTag(i->first);
765                DataTypes::ValueType::size_type ooff=optr->getOffsetForTag(i->first);
766                for (int j=0;j<getDataPointSize();++j)
767                {
768                    if (mvec[j+moff]>0)
769                    {
770                        self[j+toff]=ovec[ooff];
771                    }
772                }
773                }
774            // now for the default value
775            for (int j=0;j<getDataPointSize();++j)
776            {
777                if (mvec[j+mptr->getDefaultOffset()]>0)
778                {
779                    self[j+tptr->getDefaultOffset()]=ovec[0];
780                }
781            }
782        }
783    
784    *this = mask1 + mask2;      return;         // ugly
785      }
786      // mixed scalar and non-scalar operation
787      if ((selfrank>0) && (otherrank==0) && (mask2.getDataPointShape()==getDataPointShape()))
788      {
789            size_t num_points=self.size();
790        // OPENMP 3.0 allows unsigned loop vars.
791        #if defined(_OPENMP) && (_OPENMP < 200805)
792        long i;
793        #else
794        size_t i;
795        #endif  
796        size_t psize=getDataPointSize();    
797        #pragma omp parallel for private(i) schedule(static)
798        for (i=0;i<num_points;++i)
799        {
800            if (mvec[i]>0)
801            {
802                self[i]=ovec[i/psize];      // since this is expanded there is one scalar
803            }                   // dest point
804        }
805        return;         // ugly!
806      }
807      // tagged data is already taken care of so we only need to worry about shapes
808      // special cases with scalars are already dealt with so all we need to worry about is shape
809      if ((getDataPointShape()!=other2.getDataPointShape()) || getDataPointShape()!=mask2.getDataPointShape())
810      {
811        ostringstream oss;
812        oss <<"Error - size mismatch in arguments to copyWithMask.";
813        oss << "\nself_shape=" << DataTypes::shapeToString(getDataPointShape());
814        oss << " other2_shape=" << DataTypes::shapeToString(other2.getDataPointShape());
815        oss << " mask2_shape=" << DataTypes::shapeToString(mask2.getDataPointShape());
816        throw DataException(oss.str());
817      }
818      size_t num_points=self.size();
819    
820      // OPENMP 3.0 allows unsigned loop vars.
821    #if defined(_OPENMP) && (_OPENMP < 200805)
822      long i;
823    #else
824      size_t i;
825    #endif
826      #pragma omp parallel for private(i) schedule(static)
827      for (i=0;i<num_points;++i)
828      {
829        if (mvec[i]>0)
830        {
831           self[i]=ovec[i];
832        }
833      }
834  }  }
835    
836  bool  bool
# Line 344  Data::isExpanded() const Line 841  Data::isExpanded() const
841  }  }
842    
843  bool  bool
844    Data::actsExpanded() const
845    {
846      return m_data->actsExpanded();
847    
848    }
849    
850    
851    bool
852  Data::isTagged() const  Data::isTagged() const
853  {  {
854    DataTagged* temp=dynamic_cast<DataTagged*>(m_data.get());    DataTagged* temp=dynamic_cast<DataTagged*>(m_data.get());
855    return (temp!=0);    return (temp!=0);
856  }  }
857    
 /* TODO */  
 /* global reduction -- the local data being empty does not imply that it is empty on other processers*/  
858  bool  bool
859  Data::isEmpty() const  Data::isEmpty() const
860  {  {
# Line 366  Data::isConstant() const Line 869  Data::isConstant() const
869    return (temp!=0);    return (temp!=0);
870  }  }
871    
872    bool
873    Data::isLazy() const
874    {
875      return m_lazy;    // not asking m_data because we need to be able to ask this while m_data is changing
876    }
877    
878    // at the moment this is synonymous with !isLazy() but that could change
879    bool
880    Data::isReady() const
881    {
882      return (dynamic_cast<DataReady*>(m_data.get())!=0);
883    }
884    
885    
886  void  void
887  Data::setProtection()  Data::setProtection()
888  {  {
889     m_protected=true;     m_protected=true;
890  }  }
891    
892  bool  bool
893  Data::isProtected() const  Data::isProtected() const
894  {  {
895     return m_protected;     return m_protected;
896  }  }
897    
# Line 386  Data::expand() Line 903  Data::expand()
903    if (isConstant()) {    if (isConstant()) {
904      DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());      DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());
905      DataAbstract* temp=new DataExpanded(*tempDataConst);      DataAbstract* temp=new DataExpanded(*tempDataConst);
906      shared_ptr<DataAbstract> temp_data(temp);  //     m_data=temp->getPtr();
907      m_data=temp_data;      set_m_data(temp->getPtr());
908    } else if (isTagged()) {    } else if (isTagged()) {
909      DataTagged* tempDataTag=dynamic_cast<DataTagged*>(m_data.get());      DataTagged* tempDataTag=dynamic_cast<DataTagged*>(m_data.get());
910      DataAbstract* temp=new DataExpanded(*tempDataTag);      DataAbstract* temp=new DataExpanded(*tempDataTag);
911      shared_ptr<DataAbstract> temp_data(temp);  //     m_data=temp->getPtr();
912      m_data=temp_data;      set_m_data(temp->getPtr());
913    } else if (isExpanded()) {    } else if (isExpanded()) {
914      //      //
915      // do nothing      // do nothing
916    } else if (isEmpty()) {    } else if (isEmpty()) {
917      throw DataException("Error - Expansion of DataEmpty not possible.");      throw DataException("Error - Expansion of DataEmpty not possible.");
918      } else if (isLazy()) {
919        resolve();
920        expand();       // resolve might not give us expanded data
921    } else {    } else {
922      throw DataException("Error - Expansion not implemented for this Data type.");      throw DataException("Error - Expansion not implemented for this Data type.");
923    }    }
# Line 409  Data::tag() Line 929  Data::tag()
929    if (isConstant()) {    if (isConstant()) {
930      DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());      DataConstant* tempDataConst=dynamic_cast<DataConstant*>(m_data.get());
931      DataAbstract* temp=new DataTagged(*tempDataConst);      DataAbstract* temp=new DataTagged(*tempDataConst);
932      shared_ptr<DataAbstract> temp_data(temp);  //     m_data=temp->getPtr();
933      m_data=temp_data;      set_m_data(temp->getPtr());
934    } else if (isTagged()) {    } else if (isTagged()) {
935      // do nothing      // do nothing
936    } else if (isExpanded()) {    } else if (isExpanded()) {
937      throw DataException("Error - Creating tag data from DataExpanded not possible.");      throw DataException("Error - Creating tag data from DataExpanded not possible.");
938    } else if (isEmpty()) {    } else if (isEmpty()) {
939      throw DataException("Error - Creating tag data from DataEmpty not possible.");      throw DataException("Error - Creating tag data from DataEmpty not possible.");
940      } else if (isLazy()) {
941         DataAbstract_ptr res=m_data->resolve();
942         if (m_data->isExpanded())
943         {
944        throw DataException("Error - data would resolve to DataExpanded, tagging is not possible.");
945         }
946    //      m_data=res;
947         set_m_data(res);
948         tag();
949    } else {    } else {
950      throw DataException("Error - Tagging not implemented for this Data type.");      throw DataException("Error - Tagging not implemented for this Data type.");
951    }    }
952  }  }
953    
954  void  void
955  Data::reshapeDataPoint(const DataArrayView::ShapeType& shape)  Data::resolve()
956    {
957      if (isLazy())
958      {
959    //      m_data=m_data->resolve();
960        set_m_data(m_data->resolve());
961      }
962    }
963    
964    void
965    Data::requireWrite()
966  {  {
967    m_data->reshapeDataPoint(shape);    resolve();
968      exclusiveWrite();
969    }
970    
971    Data
972    Data::oneOver() const
973    {
974      MAKELAZYOP(RECIP)
975      return C_TensorUnaryOperation(*this, bind1st(divides<double>(),1.));
976  }  }
977    
978  Data  Data
979  Data::wherePositive() const  Data::wherePositive() const
980  {  {
981  #if defined DOPROF    MAKELAZYOP(GZ)
982    profData->where++;    return C_TensorUnaryOperation(*this, bind2nd(greater<double>(),0.0));
 #endif  
   return escript::unaryOp(*this,bind2nd(greater<double>(),0.0));  
983  }  }
984    
985  Data  Data
986  Data::whereNegative() const  Data::whereNegative() const
987  {  {
988  #if defined DOPROF    MAKELAZYOP(LZ)
989    profData->where++;    return C_TensorUnaryOperation(*this, bind2nd(less<double>(),0.0));
 #endif  
   return escript::unaryOp(*this,bind2nd(less<double>(),0.0));  
990  }  }
991    
992  Data  Data
993  Data::whereNonNegative() const  Data::whereNonNegative() const
994  {  {
995  #if defined DOPROF    MAKELAZYOP(GEZ)
996    profData->where++;    return C_TensorUnaryOperation(*this, bind2nd(greater_equal<double>(),0.0));
 #endif  
   return escript::unaryOp(*this,bind2nd(greater_equal<double>(),0.0));  
997  }  }
998    
999  Data  Data
1000  Data::whereNonPositive() const  Data::whereNonPositive() const
1001  {  {
1002  #if defined DOPROF    MAKELAZYOP(LEZ)
1003    profData->where++;    return C_TensorUnaryOperation(*this, bind2nd(less_equal<double>(),0.0));
 #endif  
   return escript::unaryOp(*this,bind2nd(less_equal<double>(),0.0));  
1004  }  }
1005    
1006  Data  Data
1007  Data::whereZero(double tol) const  Data::whereZero(double tol) const
1008  {  {
1009  #if defined DOPROF  //   Data dataAbs=abs();
1010    profData->where++;  //   return C_TensorUnaryOperation(dataAbs, bind2nd(less_equal<double>(),tol));
1011  #endif     MAKELAZYOPOFF(EZ,tol)
1012    Data dataAbs=abs();     return C_TensorUnaryOperation(*this, bind2nd(AbsLTE(),tol));
1013    return escript::unaryOp(dataAbs,bind2nd(less_equal<double>(),tol));  
1014  }  }
1015    
1016  Data  Data
1017  Data::whereNonZero(double tol) const  Data::whereNonZero(double tol) const
1018  {  {
1019  #if defined DOPROF  //   Data dataAbs=abs();
1020    profData->where++;  //   return C_TensorUnaryOperation(dataAbs, bind2nd(greater<double>(),tol));
1021  #endif    MAKELAZYOPOFF(NEZ,tol)
1022    Data dataAbs=abs();    return C_TensorUnaryOperation(*this, bind2nd(AbsGT(),tol));
1023    return escript::unaryOp(dataAbs,bind2nd(greater<double>(),tol));  
1024  }  }
1025    
1026  Data  Data
1027  Data::interpolate(const FunctionSpace& functionspace) const  Data::interpolate(const FunctionSpace& functionspace) const
1028  {  {
 #if defined DOPROF  
   profData->interpolate++;  
 #endif  
1029    return Data(*this,functionspace);    return Data(*this,functionspace);
1030  }  }
1031    
1032  bool  bool
1033  Data::probeInterpolation(const FunctionSpace& functionspace) const  Data::probeInterpolation(const FunctionSpace& functionspace) const
1034  {  {
1035    if (getFunctionSpace()==functionspace) {    return getFunctionSpace().probeInterpolation(functionspace);
     return true;  
   } else {  
     const AbstractDomain& domain=getDomain();  
     if  (domain==functionspace.getDomain()) {  
       return domain.probeInterpolationOnDomain(getFunctionSpace().getTypeCode(),functionspace.getTypeCode());  
     } else {  
       return domain.probeInterpolationACross(getFunctionSpace().getTypeCode(),functionspace.getDomain(),functionspace.getTypeCode());  
     }  
   }  
1036  }  }
1037    
1038  Data  Data
1039  Data::gradOn(const FunctionSpace& functionspace) const  Data::gradOn(const FunctionSpace& functionspace) const
1040  {  {
1041  #if defined DOPROF    if (isEmpty())
1042    profData->grad++;    {
1043  #endif      throw DataException("Error - operation not permitted on instances of DataEmpty.");
1044      }
1045      double blocktimer_start = blocktimer_time();
1046    if (functionspace.getDomain()!=getDomain())    if (functionspace.getDomain()!=getDomain())
1047      throw DataException("Error - gradient cannot be calculated on different domains.");      throw DataException("Error - gradient cannot be calculated on different domains.");
1048    DataArrayView::ShapeType grad_shape=getPointDataView().getShape();    DataTypes::ShapeType grad_shape=getDataPointShape();
1049    grad_shape.push_back(functionspace.getDim());    grad_shape.push_back(functionspace.getDim());
1050    Data out(0.0,grad_shape,functionspace,true);    Data out(0.0,grad_shape,functionspace,true);
1051    getDomain().setToGradient(out,*this);    getDomain()->setToGradient(out,*this);
1052      blocktimer_increment("grad()", blocktimer_start);
1053    return out;    return out;
1054  }  }
1055    
1056  Data  Data
1057  Data::grad() const  Data::grad() const
1058  {  {
1059    return gradOn(escript::function(getDomain()));    if (isEmpty())
1060      {
1061        throw DataException("Error - operation not permitted on instances of DataEmpty.");
1062      }
1063      return gradOn(escript::function(*getDomain()));
1064  }  }
1065    
1066  int  int
1067  Data::getDataPointSize() const  Data::getDataPointSize() const
1068  {  {
1069    return getPointDataView().noValues();    return m_data->getNoValues();
1070  }  }
1071    
1072  DataArrayView::ValueType::size_type  
1073    DataTypes::ValueType::size_type
1074  Data::getLength() const  Data::getLength() const
1075  {  {
1076    return m_data->getLength();    return m_data->getLength();
1077  }  }
1078    
1079  const DataArrayView::ShapeType&  
1080  Data::getDataPointShape() const  // There is no parallelism here ... elements need to be added in the correct order.
1081    //   If we could presize the list and then fill in the elements it might work
1082    //   This would need setting elements to be threadsafe.
1083    //   Having mulitple C threads calling into one interpreter is aparently a no-no.
1084    const boost::python::object
1085    Data::toListOfTuples(bool scalarastuple)
1086    {
1087        using namespace boost::python;
1088        using boost::python::list;
1089        if (get_MPISize()>1)
1090        {
1091            throw DataException("::toListOfTuples is not available for MPI with more than one process.");
1092        }
1093        unsigned int rank=getDataPointRank();
1094        unsigned int size=getDataPointSize();
1095    
1096        int npoints=getNumDataPoints();
1097        expand();           // This will also resolve if required
1098        const DataTypes::ValueType& vec=getReady()->getVectorRO();
1099        boost::python::list temp;
1100        temp.append(object());
1101        boost::python::list res(temp*npoints);// presize the list by the "[None] * npoints"  trick
1102        if (rank==0)
1103        {
1104            long count;
1105            if (scalarastuple)
1106            {
1107                for (count=0;count<npoints;++count)
1108                {
1109            res[count]=make_tuple(vec[count]);
1110                }
1111            }
1112            else
1113            {
1114                for (count=0;count<npoints;++count)
1115                {
1116                    res[count]=vec[count];
1117                }
1118            }
1119        }
1120        else if (rank==1)
1121        {
1122            size_t count;
1123            size_t offset=0;
1124            for (count=0;count<npoints;++count,offset+=size)
1125            {
1126                res[count]=pointToTuple1(getDataPointShape(), vec, offset);
1127            }
1128        }
1129        else if (rank==2)
1130        {
1131            size_t count;
1132            size_t offset=0;
1133            for (count=0;count<npoints;++count,offset+=size)
1134            {
1135            res[count]=pointToTuple2(getDataPointShape(), vec, offset);
1136            }
1137        }
1138        else if (rank==3)
1139        {
1140            size_t count;
1141            size_t offset=0;
1142            for (count=0;count<npoints;++count,offset+=size)
1143            {
1144                res[count]=pointToTuple3(getDataPointShape(), vec, offset);
1145            }
1146        }
1147        else if (rank==4)
1148        {
1149            size_t count;
1150            size_t offset=0;
1151            for (count=0;count<npoints;++count,offset+=size)
1152            {
1153                res[count]=pointToTuple4(getDataPointShape(), vec, offset);
1154            }
1155        }
1156        else
1157        {
1158            throw DataException("Unknown rank in ::toListOfTuples()");
1159        }
1160        return res;
1161    }
1162    
1163    const boost::python::object
1164    Data::getValueOfDataPointAsTuple(int dataPointNo)
1165    {
1166       forceResolve();
1167       if (getNumDataPointsPerSample()>0) {
1168           int sampleNo = dataPointNo/getNumDataPointsPerSample();
1169           int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
1170           //
1171           // Check a valid sample number has been supplied
1172           if ((sampleNo >= getNumSamples()) || (sampleNo < 0 )) {
1173               throw DataException("Error - Data::getValueOfDataPointAsTuple: invalid sampleNo.");
1174           }
1175    
1176           //
1177           // Check a valid data point number has been supplied
1178           if ((dataPointNoInSample >= getNumDataPointsPerSample()) || (dataPointNoInSample < 0)) {
1179               throw DataException("Error - Data::getValueOfDataPointAsTuple: invalid dataPointNoInSample.");
1180           }
1181           // TODO: global error handling
1182           DataTypes::ValueType::size_type offset=getDataOffset(sampleNo, dataPointNoInSample);
1183           return pointToTuple(getDataPointShape(),&(getDataAtOffsetRO(offset)));
1184      }
1185      else
1186      {
1187        // The pre-numpy method would return an empty array of the given shape
1188        // I'm going to throw an exception because if we have zero points per sample we have problems
1189        throw DataException("Error - need at least 1 datapoint per sample.");
1190      }
1191    
1192    }
1193    
1194    
1195    void
1196    Data::setValueOfDataPointToPyObject(int dataPointNo, const boost::python::object& py_object)
1197  {  {
1198    return getPointDataView().getShape();      // this will throw if the value cannot be represented
1199        setValueOfDataPointToArray(dataPointNo,py_object);
1200  }  }
1201    
1202  void  void
1203  Data::fillFromNumArray(const boost::python::numeric::array num_array)  Data::setValueOfDataPointToArray(int dataPointNo, const boost::python::object& obj)
1204  {  {
1205    if (isProtected()) {    if (isProtected()) {
1206          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
1207    }    }
1208    
1209      WrappedArray w(obj);
1210    //    //
1211    // check rank    // check rank
1212    if (num_array.getrank()<getDataPointRank())    if (static_cast<unsigned int>(w.getRank())<getDataPointRank())
1213        throw DataException("Rank of numarray does not match Data object rank");        throw DataException("Rank of array does not match Data object rank");
1214    
1215    //    //
1216    // check shape of num_array    // check shape of array
1217    for (int i=0; i<getDataPointRank(); i++) {    for (unsigned int i=0; i<getDataPointRank(); i++) {
1218      if (extract<int>(num_array.getshape()[i+1])!=getDataPointShape()[i])      if (w.getShape()[i]!=getDataPointShape()[i])
1219         throw DataException("Shape of numarray does not match Data object rank");         throw DataException("Shape of array does not match Data object rank");
1220    }    }
1221    
1222      exclusiveWrite();
1223    
1224    //    //
1225    // make sure data is expanded:    // make sure data is expanded:
1226      //
1227    if (!isExpanded()) {    if (!isExpanded()) {
1228      expand();      expand();
1229    }    }
1230      if (getNumDataPointsPerSample()>0) {
1231    //         int sampleNo = dataPointNo/getNumDataPointsPerSample();
1232    // and copy over         int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
1233    m_data->copyAll(num_array);         m_data->copyToDataPoint(sampleNo, dataPointNoInSample,w);
1234      } else {
1235           m_data->copyToDataPoint(-1, 0,w);
1236      }
1237  }  }
1238    
1239  const  void
1240  boost::python::numeric::array  Data::setValueOfDataPoint(int dataPointNo, const double value)
 Data::convertToNumArray()  
1241  {  {
1242    //    if (isProtected()) {
1243    // determine the total number of data points          throw DataException("Error - attempt to update protected Data object.");
   int numSamples = getNumSamples();  
   int numDataPointsPerSample = getNumDataPointsPerSample();  
   int numDataPoints = numSamples * numDataPointsPerSample;  
   
   //  
   // determine the rank and shape of each data point  
   int dataPointRank = getDataPointRank();  
   DataArrayView::ShapeType dataPointShape = getDataPointShape();  
   
   //  
   // create the numeric array to be returned  
   boost::python::numeric::array numArray(0.0);  
   
   //  
   // the rank of the returned numeric array will be the rank of  
   // the data points, plus one. Where the rank of the array is n,  
   // the last n-1 dimensions will be equal to the shape of the  
   // data points, whilst the first dimension will be equal to the  
   // total number of data points. Thus the array will consist of  
   // a serial vector of the data points.  
   int arrayRank = dataPointRank + 1;  
   DataArrayView::ShapeType arrayShape;  
   arrayShape.push_back(numDataPoints);  
   for (int d=0; d<dataPointRank; d++) {  
      arrayShape.push_back(dataPointShape[d]);  
1244    }    }
   
1245    //    //
1246    // resize the numeric array to the shape just calculated    // make sure data is expanded:
1247    if (arrayRank==1) {    exclusiveWrite();
1248      numArray.resize(arrayShape[0]);    if (!isExpanded()) {
1249    }      expand();
   if (arrayRank==2) {  
     numArray.resize(arrayShape[0],arrayShape[1]);  
   }  
   if (arrayRank==3) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2]);  
   }  
   if (arrayRank==4) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3]);  
   }  
   if (arrayRank==5) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3],arrayShape[4]);  
1250    }    }
1251      if (getNumDataPointsPerSample()>0) {
1252    //         int sampleNo = dataPointNo/getNumDataPointsPerSample();
1253    // loop through each data point in turn, loading the values for that data point         int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
1254    // into the numeric array.         m_data->copyToDataPoint(sampleNo, dataPointNoInSample,value);
1255    int dataPoint = 0;    } else {
1256    for (int sampleNo = 0; sampleNo < numSamples; sampleNo++) {         m_data->copyToDataPoint(-1, 0,value);
     for (int dataPointNo = 0; dataPointNo < numDataPointsPerSample; dataPointNo++) {  
       DataArrayView dataPointView = getDataPoint(sampleNo, dataPointNo);  
       if (dataPointRank==0) {  
         numArray[dataPoint]=dataPointView();  
       }  
       if (dataPointRank==1) {  
         for (int i=0; i<dataPointShape[0]; i++) {  
           numArray[dataPoint][i]=dataPointView(i);  
         }  
       }  
       if (dataPointRank==2) {  
         for (int i=0; i<dataPointShape[0]; i++) {  
           for (int j=0; j<dataPointShape[1]; j++) {  
             numArray[dataPoint][i][j] = dataPointView(i,j);  
           }  
         }  
       }  
       if (dataPointRank==3) {  
         for (int i=0; i<dataPointShape[0]; i++) {  
           for (int j=0; j<dataPointShape[1]; j++) {  
             for (int k=0; k<dataPointShape[2]; k++) {  
               numArray[dataPoint][i][j][k]=dataPointView(i,j,k);  
             }  
           }  
         }  
       }  
       if (dataPointRank==4) {  
         for (int i=0; i<dataPointShape[0]; i++) {  
           for (int j=0; j<dataPointShape[1]; j++) {  
             for (int k=0; k<dataPointShape[2]; k++) {  
               for (int l=0; l<dataPointShape[3]; l++) {  
                 numArray[dataPoint][i][j][k][l]=dataPointView(i,j,k,l);  
               }  
             }  
           }  
         }  
       }  
       dataPoint++;  
     }  
1257    }    }
   
   //  
   // return the loaded array  
   return numArray;  
1258  }  }
1259    
1260  const  const
1261  boost::python::numeric::array  boost::python::object
1262  Data::convertToNumArrayFromSampleNo(int sampleNo)  Data::getValueOfGlobalDataPointAsTuple(int procNo, int dataPointNo)
1263  {  {
1264    //    // This could be lazier than it is now
1265    // Check a valid sample number has been supplied    forceResolve();
   if (sampleNo >= getNumSamples()) {  
     throw DataException("Error - Data::convertToNumArray: invalid sampleNo.");  
   }  
   
   //  
   // determine the number of data points per sample  
   int numDataPointsPerSample = getNumDataPointsPerSample();  
   
   //  
   // determine the rank and shape of each data point  
   int dataPointRank = getDataPointRank();  
   DataArrayView::ShapeType dataPointShape = getDataPointShape();  
   
   //  
   // create the numeric array to be returned  
   boost::python::numeric::array numArray(0.0);  
1266    
1267    //    // copy datapoint into a buffer
1268    // the rank of the returned numeric array will be the rank of    // broadcast buffer to all nodes
1269    // the data points, plus one. Where the rank of the array is n,    // convert buffer to tuple
1270    // the last n-1 dimensions will be equal to the shape of the    // return tuple
1271    // data points, whilst the first dimension will be equal to the  
1272    // total number of data points. Thus the array will consist of    const DataTypes::ShapeType& dataPointShape = getDataPointShape();
1273    // a serial vector of the data points.    size_t length=DataTypes::noValues(dataPointShape);
1274    int arrayRank = dataPointRank + 1;  
1275    DataArrayView::ShapeType arrayShape;    // added for the MPI communication
1276    arrayShape.push_back(numDataPointsPerSample);    double *tmpData = new double[length];
1277    for (int d=0; d<dataPointRank; d++) {  
1278       arrayShape.push_back(dataPointShape[d]);    // updated for the MPI case
1279    }    if( get_MPIRank()==procNo ){
1280          if (getNumDataPointsPerSample()>0) {
1281        int sampleNo = dataPointNo/getNumDataPointsPerSample();
1282        int dataPointNoInSample = dataPointNo - sampleNo * getNumDataPointsPerSample();
1283        //
1284        // Check a valid sample number has been supplied
1285        if ((sampleNo >= getNumSamples()) || (sampleNo < 0 )) {
1286            throw DataException("Error - Data::getValueOfGlobalDataPointAsTuple: invalid sampleNo.");
1287        }
1288    
1289    //      //
1290    // resize the numeric array to the shape just calculated      // Check a valid data point number has been supplied
1291    if (arrayRank==1) {      if ((dataPointNoInSample >= getNumDataPointsPerSample()) || (dataPointNoInSample < 0)) {
1292      numArray.resize(arrayShape[0]);          throw DataException("Error - Data::getValueOfGlobalDataPointAsTuple: invalid dataPointNoInSample.");
1293    }      }
1294    if (arrayRank==2) {      // TODO: global error handling
1295      numArray.resize(arrayShape[0],arrayShape[1]);      DataTypes::ValueType::size_type offset=getDataOffset(sampleNo, dataPointNoInSample);
   }  
   if (arrayRank==3) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2]);  
   }  
   if (arrayRank==4) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3]);  
   }  
   if (arrayRank==5) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3],arrayShape[4]);  
   }  
1296    
1297    //      memcpy(tmpData,&(getDataAtOffsetRO(offset)),length*sizeof(double));
1298    // loop through each data point in turn, loading the values for that data point       }
   // into the numeric array.  
   for (int dataPoint = 0; dataPoint < numDataPointsPerSample; dataPoint++) {  
     DataArrayView dataPointView = getDataPoint(sampleNo, dataPoint);  
     if (dataPointRank==0) {  
       numArray[dataPoint]=dataPointView();  
     }  
     if (dataPointRank==1) {  
       for (int i=0; i<dataPointShape[0]; i++) {  
         numArray[dataPoint][i]=dataPointView(i);  
       }  
     }  
     if (dataPointRank==2) {  
       for (int i=0; i<dataPointShape[0]; i++) {  
         for (int j=0; j<dataPointShape[1]; j++) {  
           numArray[dataPoint][i][j] = dataPointView(i,j);  
         }  
       }  
     }  
     if (dataPointRank==3) {  
       for (int i=0; i<dataPointShape[0]; i++) {  
         for (int j=0; j<dataPointShape[1]; j++) {  
           for (int k=0; k<dataPointShape[2]; k++) {  
             numArray[dataPoint][i][j][k]=dataPointView(i,j,k);  
           }  
         }  
       }  
     }  
     if (dataPointRank==4) {  
       for (int i=0; i<dataPointShape[0]; i++) {  
         for (int j=0; j<dataPointShape[1]; j++) {  
           for (int k=0; k<dataPointShape[2]; k++) {  
             for (int l=0; l<dataPointShape[3]; l++) {  
               numArray[dataPoint][i][j][k][l]=dataPointView(i,j,k,l);  
             }  
           }  
         }  
       }  
     }  
1299    }    }
1300    #ifdef PASO_MPI
1301      // broadcast the data to all other processes
1302      MPI_Bcast( tmpData, length, MPI_DOUBLE, procNo, get_MPIComm() );
1303    #endif
1304    
1305      boost::python::tuple t=pointToTuple(dataPointShape,tmpData);
1306      delete [] tmpData;
1307    //    //
1308    // return the loaded array    // return the loaded array
1309    return numArray;    return t;
1310    
1311  }  }
1312    
 const  
 boost::python::numeric::array  
 Data::convertToNumArrayFromDPNo(int procNo,  
                                 int sampleNo,  
                                                                 int dataPointNo)  
1313    
1314    boost::python::object
1315    Data::integrateToTuple_const() const
1316  {  {
1317      size_t length=0;    if (isLazy())
1318      int i, j, k, l, pos;    {
1319        throw DataException("Error - cannot integrate for constant lazy data.");
   //  
   // Check a valid sample number has been supplied  
   if (sampleNo >= getNumSamples()) {  
     throw DataException("Error - Data::convertToNumArray: invalid sampleNo.");  
   }  
   
   //  
   // Check a valid data point number has been supplied  
   if (dataPointNo >= getNumDataPointsPerSample()) {  
     throw DataException("Error - Data::convertToNumArray: invalid dataPointNo.");  
   }  
   
   //  
   // determine the rank and shape of each data point  
   int dataPointRank = getDataPointRank();  
   DataArrayView::ShapeType dataPointShape = getDataPointShape();  
   
   //  
   // create the numeric array to be returned  
   boost::python::numeric::array numArray(0.0);  
   
   //  
   // the shape of the returned numeric array will be the same  
   // as that of the data point  
   int arrayRank = dataPointRank;  
   DataArrayView::ShapeType arrayShape = dataPointShape;  
   
   //  
   // resize the numeric array to the shape just calculated  
   if (arrayRank==0) {  
     numArray.resize(1);  
   }  
   if (arrayRank==1) {  
     numArray.resize(arrayShape[0]);  
   }  
   if (arrayRank==2) {  
     numArray.resize(arrayShape[0],arrayShape[1]);  
   }  
   if (arrayRank==3) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2]);  
   }  
   if (arrayRank==4) {  
     numArray.resize(arrayShape[0],arrayShape[1],arrayShape[2],arrayShape[3]);  
1320    }    }
1321      return integrateWorker();
1322    }
1323    
1324      // added for the MPI communication  boost::python::object
1325      length=1;  Data::integrateToTuple()
1326      for( i=0; i<arrayRank; i++ )  {
1327          length *= arrayShape[i];    if (isLazy())
1328      double *tmpData = new double[length];    {
1329        expand();   // Can't do a non-resolving version of this without changing the domain object
1330    //    }         // see the dom->setToIntegrals call. Not saying it can't be done, just not doing it yet.
1331    // load the values for the data point into the numeric array.    return integrateWorker();
   
     // updated for the MPI case  
     if( get_MPIRank()==procNo ){  
         // create a view of the data if it is stored locally  
         DataArrayView dataPointView = getDataPoint(sampleNo, dataPointNo);  
           
         // pack the data from the view into tmpData for MPI communication  
         pos=0;  
         switch( dataPointRank ){  
             case 0 :  
                 tmpData[0] = dataPointView();  
                 break;  
             case 1 :          
                 for( i=0; i<dataPointShape[0]; i++ )  
                     tmpData[i]=dataPointView(i);  
                 break;  
             case 2 :          
                 for( i=0; i<dataPointShape[0]; i++ )  
                     for( j=0; j<dataPointShape[1]; j++, pos++ )  
                         tmpData[pos]=dataPointView(i,j);  
                 break;  
             case 3 :          
                 for( i=0; i<dataPointShape[0]; i++ )  
                     for( j=0; j<dataPointShape[1]; j++ )  
                         for( k=0; k<dataPointShape[2]; k++, pos++ )  
                             tmpData[pos]=dataPointView(i,j,k);  
                 break;  
             case 4 :  
                 for( i=0; i<dataPointShape[0]; i++ )  
                     for( j=0; j<dataPointShape[1]; j++ )  
                         for( k=0; k<dataPointShape[2]; k++ )  
                             for( l=0; l<dataPointShape[3]; l++, pos++ )  
                                 tmpData[pos]=dataPointView(i,j,k,l);  
                 break;  
         }  
     }  
 #ifdef PASO_MPI  
         // broadcast the data to all other processes  
         MPI_Bcast( tmpData, length, MPI_DOUBLE, procNo, get_MPIComm() );  
 #endif  
   
     // unpack the data  
     switch( dataPointRank ){  
         case 0 :  
             numArray[i]=tmpData[0];  
             break;  
         case 1 :          
             for( i=0; i<dataPointShape[0]; i++ )  
                 numArray[i]=tmpData[i];  
             break;  
         case 2 :          
             for( i=0; i<dataPointShape[0]; i++ )  
                 for( j=0; j<dataPointShape[1]; j++ )  
                     tmpData[i+j*dataPointShape[0]];  
             break;  
         case 3 :          
             for( i=0; i<dataPointShape[0]; i++ )  
                 for( j=0; j<dataPointShape[1]; j++ )  
                     for( k=0; k<dataPointShape[2]; k++ )  
                         tmpData[i+dataPointShape[0]*(j*+k*dataPointShape[1])];  
             break;  
         case 4 :  
             for( i=0; i<dataPointShape[0]; i++ )  
                 for( j=0; j<dataPointShape[1]; j++ )  
                     for( k=0; k<dataPointShape[2]; k++ )  
                         for( l=0; l<dataPointShape[3]; l++ )  
                             tmpData[i+dataPointShape[0]*(j*+dataPointShape[1]*(k+l*dataPointShape[2]))];  
             break;  
     }  
   
     delete [] tmpData;    
 /*  
   if (dataPointRank==0) {  
     numArray[0]=dataPointView();  
   }  
   if (dataPointRank==1) {  
     for (int i=0; i<dataPointShape[0]; i++) {  
       numArray[i]=dataPointView(i);  
     }  
   }  
   if (dataPointRank==2) {  
     for (int i=0; i<dataPointShape[0]; i++) {  
       for (int j=0; j<dataPointShape[1]; j++) {  
         numArray[i][j] = dataPointView(i,j);  
       }  
     }  
   }  
   if (dataPointRank==3) {  
     for (int i=0; i<dataPointShape[0]; i++) {  
       for (int j=0; j<dataPointShape[1]; j++) {  
         for (int k=0; k<dataPointShape[2]; k++) {  
           numArray[i][j][k]=dataPointView(i,j,k);  
         }  
       }  
     }  
   }  
   if (dataPointRank==4) {  
     for (int i=0; i<dataPointShape[0]; i++) {  
       for (int j=0; j<dataPointShape[1]; j++) {  
         for (int k=0; k<dataPointShape[2]; k++) {  
           for (int l=0; l<dataPointShape[3]; l++) {  
             numArray[i][j][k][l]=dataPointView(i,j,k,l);  
           }  
         }  
       }  
     }  
   }  
 */  
1332    
   //  
   // return the loaded array  
   return numArray;  
1333  }  }
1334    
1335  boost::python::numeric::array  boost::python::object
1336  Data::integrate() const  Data::integrateWorker() const
1337  {  {
1338    int index;    DataTypes::ShapeType shape = getDataPointShape();
1339    int rank = getDataPointRank();    int dataPointSize = getDataPointSize();
   DataArrayView::ShapeType shape = getDataPointShape();  
   
 #if defined DOPROF  
   profData->integrate++;  
 #endif  
1340    
1341    //    //
1342    // calculate the integral values    // calculate the integral values
1343    vector<double> integrals(getDataPointSize());    vector<double> integrals(dataPointSize);
1344    AbstractContinuousDomain::asAbstractContinuousDomain(getDomain()).setToIntegrals(integrals,*this);    vector<double> integrals_local(dataPointSize);
1345      const AbstractContinuousDomain* dom=dynamic_cast<const AbstractContinuousDomain*>(getDomain().get());
1346    //    if (dom==0)
1347    // create the numeric array to be returned    {            
1348    // and load the array with the integral values      throw DataException("Can not integrate over non-continuous domains.");
   boost::python::numeric::array bp_array(1.0);  
   if (rank==0) {  
     bp_array.resize(1);  
     index = 0;  
     bp_array[0] = integrals[index];  
   }  
   if (rank==1) {  
     bp_array.resize(shape[0]);  
     for (int i=0; i<shape[0]; i++) {  
       index = i;  
       bp_array[i] = integrals[index];  
     }  
   }  
   if (rank==2) {  
        bp_array.resize(shape[0],shape[1]);  
        for (int i=0; i<shape[0]; i++) {  
          for (int j=0; j<shape[1]; j++) {  
            index = i + shape[0] * j;  
            bp_array[make_tuple(i,j)] = integrals[index];  
          }  
        }  
   }  
   if (rank==3) {  
     bp_array.resize(shape[0],shape[1],shape[2]);  
     for (int i=0; i<shape[0]; i++) {  
       for (int j=0; j<shape[1]; j++) {  
         for (int k=0; k<shape[2]; k++) {  
           index = i + shape[0] * ( j + shape[1] * k );  
           bp_array[make_tuple(i,j,k)] = integrals[index];  
         }  
       }  
     }  
   }  
   if (rank==4) {  
     bp_array.resize(shape[0],shape[1],shape[2],shape[3]);  
     for (int i=0; i<shape[0]; i++) {  
       for (int j=0; j<shape[1]; j++) {  
         for (int k=0; k<shape[2]; k++) {  
           for (int l=0; l<shape[3]; l++) {  
             index = i + shape[0] * ( j + shape[1] * ( k + shape[2] * l ) );  
             bp_array[make_tuple(i,j,k,l)] = integrals[index];  
           }  
         }  
       }  
     }  
1349    }    }
1350    #ifdef PASO_MPI
1351      dom->setToIntegrals(integrals_local,*this);
1352      // Global sum: use an array instead of a vector because elements of array are guaranteed to be contiguous in memory
1353      double *tmp = new double[dataPointSize];
1354      double *tmp_local = new double[dataPointSize];
1355      for (int i=0; i<dataPointSize; i++) { tmp_local[i] = integrals_local[i]; }
1356      MPI_Allreduce( &tmp_local[0], &tmp[0], dataPointSize, MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD );
1357      for (int i=0; i<dataPointSize; i++) { integrals[i] = tmp[i]; }
1358      tuple result=pointToTuple(shape,tmp);
1359      delete[] tmp;
1360      delete[] tmp_local;
1361    #else
1362      dom->setToIntegrals(integrals,*this);
1363    /*  double *tmp = new double[dataPointSize];
1364      for (int i=0; i<dataPointSize; i++) { tmp[i]=integrals[i]; }*/
1365      tuple result=pointToTuple(shape,integrals);
1366    //   delete tmp;
1367    #endif
1368    
1369    //  
1370    // return the loaded array    return result;
   return bp_array;  
1371  }  }
1372    
1373  Data  Data
1374  Data::sin() const  Data::sin() const
1375  {  {
1376  #if defined DOPROF    MAKELAZYOP(SIN)
1377    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::sin);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sin);  
1378  }  }
1379    
1380  Data  Data
1381  Data::cos() const  Data::cos() const
1382  {  {
1383  #if defined DOPROF    MAKELAZYOP(COS)
1384    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::cos);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::cos);  
1385  }  }
1386    
1387  Data  Data
1388  Data::tan() const  Data::tan() const
1389  {  {
1390  #if defined DOPROF    MAKELAZYOP(TAN)
1391    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::tan);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::tan);  
1392  }  }
1393    
1394  Data  Data
1395  Data::asin() const  Data::asin() const
1396  {  {
1397  #if defined DOPROF    MAKELAZYOP(ASIN)
1398    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::asin);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::asin);  
1399  }  }
1400    
1401  Data  Data
1402  Data::acos() const  Data::acos() const
1403  {  {
1404  #if defined DOPROF    MAKELAZYOP(ACOS)
1405    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::acos);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::acos);  
1406  }  }
1407    
1408    
1409  Data  Data
1410  Data::atan() const  Data::atan() const
1411  {  {
1412  #if defined DOPROF    MAKELAZYOP(ATAN)
1413    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::atan);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::atan);  
1414  }  }
1415    
1416  Data  Data
1417  Data::sinh() const  Data::sinh() const
1418  {  {
1419  #if defined DOPROF    MAKELAZYOP(SINH)
1420    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::sinh);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sinh);  
1421  }  }
1422    
1423  Data  Data
1424  Data::cosh() const  Data::cosh() const
1425  {  {
1426  #if defined DOPROF    MAKELAZYOP(COSH)
1427    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::cosh);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::cosh);  
1428  }  }
1429    
1430  Data  Data
1431  Data::tanh() const  Data::tanh() const
1432  {  {
1433  #if defined DOPROF    MAKELAZYOP(TANH)
1434    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::tanh);
1435    }
1436    
1437    
1438    Data
1439    Data::erf() const
1440    {
1441    #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1442      throw DataException("Error - Data:: erf function is not supported on _WIN32 platforms.");
1443    #else
1444      MAKELAZYOP(ERF)
1445      return C_TensorUnaryOperation(*this, ::erf);
1446  #endif  #endif
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::tanh);  
1447  }  }
1448    
1449  Data  Data
1450  Data::asinh() const  Data::asinh() const
1451  {  {
1452  #if defined DOPROF    MAKELAZYOP(ASINH)
1453    profData->unary++;  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1454      return C_TensorUnaryOperation(*this, escript::asinh_substitute);
1455    #else
1456      return C_TensorUnaryOperation(*this, ::asinh);
1457  #endif  #endif
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::asinh);  
1458  }  }
1459    
1460  Data  Data
1461  Data::acosh() const  Data::acosh() const
1462  {  {
1463  #if defined DOPROF    MAKELAZYOP(ACOSH)
1464    profData->unary++;  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1465      return C_TensorUnaryOperation(*this, escript::acosh_substitute);
1466    #else
1467      return C_TensorUnaryOperation(*this, ::acosh);
1468  #endif  #endif
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::acosh);  
1469  }  }
1470    
1471  Data  Data
1472  Data::atanh() const  Data::atanh() const
1473  {  {
1474  #if defined DOPROF    MAKELAZYOP(ATANH)
1475    profData->unary++;  #if defined (_WIN32) && !defined(__INTEL_COMPILER)
1476      return C_TensorUnaryOperation(*this, escript::atanh_substitute);
1477    #else
1478      return C_TensorUnaryOperation(*this, ::atanh);
1479  #endif  #endif
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::atanh);  
1480  }  }
1481    
1482  Data  Data
1483  Data::log10() const  Data::log10() const
1484  {  {
1485  #if defined DOPROF    MAKELAZYOP(LOG10)
1486    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::log10);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::log10);  
1487  }  }
1488    
1489  Data  Data
1490  Data::log() const  Data::log() const
1491  {  {
1492  #if defined DOPROF    MAKELAZYOP(LOG)
1493    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::log);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::log);  
1494  }  }
1495    
1496  Data  Data
1497  Data::sign() const  Data::sign() const
1498  {  {
1499  #if defined DOPROF    MAKELAZYOP(SIGN)
1500    profData->unary++;    return C_TensorUnaryOperation(*this, escript::fsign);
 #endif  
   return escript::unaryOp(*this,escript::fsign);  
1501  }  }
1502    
1503  Data  Data
1504  Data::abs() const  Data::abs() const
1505  {  {
1506  #if defined DOPROF    MAKELAZYOP(ABS)
1507    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::fabs);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::fabs);  
1508  }  }
1509    
1510  Data  Data
1511  Data::neg() const  Data::neg() const
1512  {  {
1513  #if defined DOPROF    MAKELAZYOP(NEG)
1514    profData->unary++;    return C_TensorUnaryOperation(*this, negate<double>());
 #endif  
   return escript::unaryOp(*this,negate<double>());  
1515  }  }
1516    
1517  Data  Data
1518  Data::pos() const  Data::pos() const
1519  {  {
1520  #if defined DOPROF      // not doing lazy check here is deliberate.
1521    profData->unary++;      // since a deep copy of lazy data should be cheap, I'll just let it happen now
 #endif  
1522    Data result;    Data result;
1523    // perform a deep copy    // perform a deep copy
1524    result.copy(*this);    result.copy(*this);
# Line 1196  Data::pos() const Line 1528  Data::pos() const
1528  Data  Data
1529  Data::exp() const  Data::exp() const
1530  {  {
1531  #if defined DOPROF    MAKELAZYOP(EXP)
1532    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::exp);
 #endif  
   return escript::unaryOp(*this,(Data::UnaryDFunPtr)::exp);  
1533  }  }
1534    
1535  Data  Data
1536  Data::sqrt() const  Data::sqrt() const
1537  {  {
1538  #if defined DOPROF    MAKELAZYOP(SQRT)
1539    profData->unary++;    return C_TensorUnaryOperation<double (*)(double)>(*this, ::sqrt);
1540    }
1541    
1542    double
1543    Data::Lsup_const() const
1544    {
1545       if (isLazy())
1546       {
1547        throw DataException("Error - cannot compute Lsup for constant lazy data.");
1548       }
1549       return LsupWorker();
1550    }
1551    
1552    double
1553    Data::Lsup()
1554    {
1555       if (isLazy())
1556       {
1557        if (!actsExpanded() || CHECK_DO_CRES)
1558        {
1559            resolve();
1560        }
1561        else
1562        {
1563    #ifdef PASO_MPI
1564            return lazyAlgWorker<AbsMax>(0,MPI_MAX);
1565    #else
1566            return lazyAlgWorker<AbsMax>(0);
1567  #endif  #endif
1568    return escript::unaryOp(*this,(Data::UnaryDFunPtr)::sqrt);      }
1569       }
1570       return LsupWorker();
1571    }
1572    
1573    double
1574    Data::sup_const() const
1575    {
1576       if (isLazy())
1577       {
1578        throw DataException("Error - cannot compute sup for constant lazy data.");
1579       }
1580       return supWorker();
1581  }  }
1582    
1583  double  double
1584  Data::Lsup() const  Data::sup()
1585  {  {
1586    double localValue, globalValue;     if (isLazy())
1587  #if defined DOPROF     {
1588    profData->reduction1++;      if (!actsExpanded() || CHECK_DO_CRES)
1589        {
1590            resolve();
1591        }
1592        else
1593        {
1594    #ifdef PASO_MPI
1595            return lazyAlgWorker<FMax>(numeric_limits<double>::max()*-1, MPI_MAX);
1596    #else
1597            return lazyAlgWorker<FMax>(numeric_limits<double>::max()*-1);
1598  #endif  #endif
1599    //      }
1600    // set the initial absolute maximum value to zero     }
1601       return supWorker();
1602    }
1603    
1604    AbsMax abs_max_func;  double
1605    localValue = algorithm(abs_max_func,0);  Data::inf_const() const
1606    {
1607       if (isLazy())
1608       {
1609        throw DataException("Error - cannot compute inf for constant lazy data.");
1610       }
1611       return infWorker();
1612    }
1613    
1614    double
1615    Data::inf()
1616    {
1617       if (isLazy())
1618       {
1619        if (!actsExpanded() || CHECK_DO_CRES)
1620        {
1621            resolve();
1622        }
1623        else
1624        {
1625  #ifdef PASO_MPI  #ifdef PASO_MPI
1626    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );          return lazyAlgWorker<FMin>(numeric_limits<double>::max(), MPI_MIN);
   return globalValue;  
1627  #else  #else
1628    return localValue;          return lazyAlgWorker<FMin>(numeric_limits<double>::max());
1629  #endif  #endif
1630        }
1631       }
1632       return infWorker();
1633  }  }
1634    
1635    template <class BinaryOp>
1636  double  double
1637  Data::Linf() const  #ifdef PASO_MPI
1638    Data::lazyAlgWorker(double init, MPI_Op mpiop_type)
1639    #else
1640    Data::lazyAlgWorker(double init)
1641    #endif
1642  {  {
1643    double localValue, globalValue;     if (!isLazy() || !m_data->actsExpanded())
1644  #if defined DOPROF     {
1645    profData->reduction1++;      throw DataException("Error - lazyAlgWorker can only be called on lazy(expanded) data.");
1646       }
1647       DataLazy* dl=dynamic_cast<DataLazy*>(m_data.get());
1648       EsysAssert((dl!=0), "Programming error - casting to DataLazy.");
1649       BufferGroup* bg=allocSampleBuffer();
1650       double val=init;
1651       int i=0;
1652       const size_t numsamples=getNumSamples();
1653       const size_t samplesize=getNoValues()*getNumDataPointsPerSample();
1654       BinaryOp operation;
1655       #pragma omp parallel private(i)
1656       {
1657        double localtot=init;
1658        #pragma omp for schedule(static) private(i)
1659        for (i=0;i<numsamples;++i)
1660        {
1661            size_t roffset=0;
1662            const DataTypes::ValueType* v=dl->resolveSample(*bg, i, roffset);
1663            // Now we have the sample, run operation on all points
1664            for (size_t j=0;j<samplesize;++j)
1665            {
1666            localtot=operation(localtot,(*v)[j+roffset]);
1667            }
1668        }
1669        #pragma omp critical
1670        val=operation(val,localtot);
1671       }
1672       freeSampleBuffer(bg);
1673    #ifdef PASO_MPI
1674       double globalValue;
1675       MPI_Allreduce( &val, &globalValue, 1, MPI_DOUBLE, mpiop_type, MPI_COMM_WORLD );
1676       return globalValue;
1677    #else
1678       return val;
1679  #endif  #endif
1680    }
1681    
1682    double
1683    Data::LsupWorker() const
1684    {
1685      double localValue;
1686    //    //
1687    // set the initial absolute minimum value to max double    // set the initial absolute maximum value to zero
   AbsMin abs_min_func;  
   localValue = algorithm(abs_min_func,numeric_limits<double>::max());  
1688    
1689      AbsMax abs_max_func;
1690      localValue = algorithm(abs_max_func,0);
1691  #ifdef PASO_MPI  #ifdef PASO_MPI
1692    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MIN, MPI_COMM_WORLD );    double globalValue;
1693      MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );
1694    return globalValue;    return globalValue;
1695  #else  #else
1696    return localValue;    return localValue;
# Line 1252  Data::Linf() const Line 1698  Data::Linf() const
1698  }  }
1699    
1700  double  double
1701  Data::sup() const  Data::supWorker() const
1702  {  {
1703    double localValue, globalValue;    double localValue;
 #if defined DOPROF  
   profData->reduction1++;  
 #endif  
1704    //    //
1705    // set the initial maximum value to min possible double    // set the initial maximum value to min possible double
1706    FMax fmax_func;    FMax fmax_func;
1707    localValue = algorithm(fmax_func,numeric_limits<double>::max()*-1);    localValue = algorithm(fmax_func,numeric_limits<double>::max()*-1);
1708  #ifdef PASO_MPI  #ifdef PASO_MPI
1709      double globalValue;
1710    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MAX, MPI_COMM_WORLD );
1711    return globalValue;    return globalValue;
1712  #else  #else
# Line 1271  Data::sup() const Line 1715  Data::sup() const
1715  }  }
1716    
1717  double  double
1718  Data::inf() const  Data::infWorker() const
1719  {  {
1720    double localValue, globalValue;    double localValue;
 #if defined DOPROF  
   profData->reduction1++;  
 #endif  
1721    //    //
1722    // set the initial minimum value to max possible double    // set the initial minimum value to max possible double
1723    FMin fmin_func;    FMin fmin_func;
1724    localValue = algorithm(fmin_func,numeric_limits<double>::max());    localValue = algorithm(fmin_func,numeric_limits<double>::max());
1725  #ifdef PASO_MPI  #ifdef PASO_MPI
1726      double globalValue;
1727    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MIN, MPI_COMM_WORLD );    MPI_Allreduce( &localValue, &globalValue, 1, MPI_DOUBLE, MPI_MIN, MPI_COMM_WORLD );
1728    return globalValue;    return globalValue;
1729  #else  #else
# Line 1289  Data::inf() const Line 1731  Data::inf() const
1731  #endif  #endif
1732  }  }
1733    
 /* TODO */  
1734  /* global reduction */  /* global reduction */
1735  Data  
1736  Data::maxval() const  
1737    inline Data
1738    Data::minval_nonlazy() const
1739    {
1740      //
1741      // set the initial minimum value to max possible double
1742      FMin fmin_func;
1743      return dp_algorithm(fmin_func,numeric_limits<double>::max());
1744    }
1745    
1746    
1747    inline Data
1748    Data::maxval_nonlazy() const
1749  {  {
 #if defined DOPROF  
   profData->reduction2++;  
 #endif  
1750    //    //
1751    // set the initial maximum value to min possible double    // set the initial maximum value to min possible double
1752    FMax fmax_func;    FMax fmax_func;
1753    return dp_algorithm(fmax_func,numeric_limits<double>::max()*-1);    return dp_algorithm(fmax_func,numeric_limits<double>::max()*-1);
1754  }  }
1755    
1756    
1757    
1758    Data
1759    Data::maxval() const
1760    {
1761       MAKELAZYOP(MAXVAL)
1762       return maxval_nonlazy();
1763    }
1764    
1765    
1766  Data  Data
1767  Data::minval() const  Data::minval() const
1768  {  {
1769  #if defined DOPROF    MAKELAZYOP(MINVAL)
1770    profData->reduction2++;    return minval_nonlazy();
 #endif  
   //  
   // set the initial minimum value to max possible double  
   FMin fmin_func;  
   return dp_algorithm(fmin_func,numeric_limits<double>::max());  
1771  }  }
1772    
1773    
1774  Data  Data
1775  Data::trace() const  Data::swapaxes(const int axis0, const int axis1) const
1776  {  {
1777  #if defined DOPROF       int axis0_tmp,axis1_tmp;
1778    profData->reduction2++;       DataTypes::ShapeType s=getDataPointShape();
1779  #endif       DataTypes::ShapeType ev_shape;
1780    Trace trace_func;       // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]
1781    return dp_algorithm(trace_func,0);       // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)
1782         int rank=getDataPointRank();
1783         if (rank<2) {
1784            throw DataException("Error - Data::swapaxes argument must have at least rank 2.");
1785         }
1786         if (axis0<0 || axis0>rank-1) {
1787            throw DataException("Error - Data::swapaxes: axis0 must be between 0 and rank-1=" + rank-1);
1788         }
1789         if (axis1<0 || axis1>rank-1) {
1790             throw DataException("Error - Data::swapaxes: axis1 must be between 0 and rank-1=" + rank-1);
1791         }
1792         if (axis0 == axis1) {
1793             throw DataException("Error - Data::swapaxes: axis indices must be different.");
1794         }
1795         MAKELAZYOP2(SWAP,axis0,axis1)
1796         if (axis0 > axis1)
1797         {
1798        axis0_tmp=axis1;
1799        axis1_tmp=axis0;
1800         }
1801         else
1802         {
1803        axis0_tmp=axis0;
1804        axis1_tmp=axis1;
1805         }
1806         for (int i=0; i<rank; i++)
1807         {
1808        if (i == axis0_tmp)
1809        {
1810            ev_shape.push_back(s[axis1_tmp]);
1811        }
1812        else if (i == axis1_tmp)
1813        {
1814            ev_shape.push_back(s[axis0_tmp]);
1815        }
1816        else
1817        {
1818            ev_shape.push_back(s[i]);
1819        }
1820         }
1821         Data ev(0.,ev_shape,getFunctionSpace());
1822         ev.typeMatchRight(*this);
1823         m_data->swapaxes(ev.m_data.get(), axis0_tmp, axis1_tmp);
1824         return ev;
1825  }  }
1826    
1827  Data  Data
1828  Data::symmetric() const  Data::symmetric() const
1829  {  {
      #if defined DOPROF  
         profData->unary++;  
      #endif  
1830       // check input       // check input
1831       DataArrayView::ShapeType s=getDataPointShape();       DataTypes::ShapeType s=getDataPointShape();
1832       if (getDataPointRank()==2) {       if (getDataPointRank()==2) {
1833          if(s[0] != s[1])          if(s[0] != s[1])
1834             throw DataException("Error - Data::symmetric can only be calculated for rank 2 object with equal first and second dimension.");             throw DataException("Error - Data::symmetric can only be calculated for rank 2 object with equal first and second dimension.");
1835       }       }
1836       else if (getDataPointRank()==4) {       else if (getDataPointRank()==4) {
# Line 1344  Data::symmetric() const Line 1840  Data::symmetric() const
1840       else {       else {
1841          throw DataException("Error - Data::symmetric can only be calculated for rank 2 or 4 object.");          throw DataException("Error - Data::symmetric can only be calculated for rank 2 or 4 object.");
1842       }       }
1843         MAKELAZYOP(SYM)
1844       Data ev(0.,getDataPointShape(),getFunctionSpace());       Data ev(0.,getDataPointShape(),getFunctionSpace());
1845       ev.typeMatchRight(*this);       ev.typeMatchRight(*this);
1846       m_data->symmetric(ev.m_data.get());       m_data->symmetric(ev.m_data.get());
# Line 1353  Data::symmetric() const Line 1850  Data::symmetric() const
1850  Data  Data
1851  Data::nonsymmetric() const  Data::nonsymmetric() const
1852  {  {
1853       #if defined DOPROF       MAKELAZYOP(NSYM)
         profData->unary++;  
      #endif  
1854       // check input       // check input
1855       DataArrayView::ShapeType s=getDataPointShape();       DataTypes::ShapeType s=getDataPointShape();
1856       if (getDataPointRank()==2) {       if (getDataPointRank()==2) {
1857          if(s[0] != s[1])          if(s[0] != s[1])
1858             throw DataException("Error - Data::nonsymmetric can only be calculated for rank 2 object with equal first and second dimension.");             throw DataException("Error - Data::nonsymmetric can only be calculated for rank 2 object with equal first and second dimension.");
1859          DataArrayView::ShapeType ev_shape;          DataTypes::ShapeType ev_shape;
1860          ev_shape.push_back(s[0]);          ev_shape.push_back(s[0]);
1861          ev_shape.push_back(s[1]);          ev_shape.push_back(s[1]);
1862          Data ev(0.,ev_shape,getFunctionSpace());          Data ev(0.,ev_shape,getFunctionSpace());
# Line 1372  Data::nonsymmetric() const Line 1867  Data::nonsymmetric() const
1867       else if (getDataPointRank()==4) {       else if (getDataPointRank()==4) {
1868          if(!(s[0] == s[2] && s[1] == s[3]))          if(!(s[0] == s[2] && s[1] == s[3]))
1869             throw DataException("Error - Data::nonsymmetric can only be calculated for rank 4 object with dim0==dim2 and dim1==dim3.");             throw DataException("Error - Data::nonsymmetric can only be calculated for rank 4 object with dim0==dim2 and dim1==dim3.");
1870          DataArrayView::ShapeType ev_shape;          DataTypes::ShapeType ev_shape;
1871          ev_shape.push_back(s[0]);          ev_shape.push_back(s[0]);
1872          ev_shape.push_back(s[1]);          ev_shape.push_back(s[1]);
1873          ev_shape.push_back(s[2]);          ev_shape.push_back(s[2]);
# Line 1388  Data::nonsymmetric() const Line 1883  Data::nonsymmetric() const
1883  }  }
1884    
1885  Data  Data
1886  Data::matrixtrace(int axis_offset) const  Data::trace(int axis_offset) const
1887  {  {    
1888       #if defined DOPROF       MAKELAZYOPOFF(TRACE,axis_offset)
1889          profData->unary++;       if ((axis_offset<0) || (axis_offset>getDataPointRank()))
1890       #endif       {
1891       DataArrayView::ShapeType s=getDataPointShape();      throw DataException("Error - Data::trace, axis_offset must be between 0 and rank-2 inclusive.");
1892         }
1893         DataTypes::ShapeType s=getDataPointShape();
1894       if (getDataPointRank()==2) {       if (getDataPointRank()==2) {
1895          DataArrayView::ShapeType ev_shape;          DataTypes::ShapeType ev_shape;
1896          Data ev(0.,ev_shape,getFunctionSpace());          Data ev(0.,ev_shape,getFunctionSpace());
1897          ev.typeMatchRight(*this);          ev.typeMatchRight(*this);
1898          m_data->matrixtrace(ev.m_data.get(), axis_offset);          m_data->trace(ev.m_data.get(), axis_offset);
1899          return ev;          return ev;
1900       }       }
1901       if (getDataPointRank()==3) {       if (getDataPointRank()==3) {
1902          DataArrayView::ShapeType ev_shape;          DataTypes::ShapeType ev_shape;
1903          if (axis_offset==0) {          if (axis_offset==0) {
1904            int s2=s[2];            int s2=s[2];
1905            ev_shape.push_back(s2);            ev_shape.push_back(s2);
# Line 1413  Data::matrixtrace(int axis_offset) const Line 1910  Data::matrixtrace(int axis_offset) const
1910          }          }
1911          Data ev(0.,ev_shape,getFunctionSpace());          Data ev(0.,ev_shape,getFunctionSpace());
1912          ev.typeMatchRight(*this);          ev.typeMatchRight(*this);
1913          m_data->matrixtrace(ev.m_data.get(), axis_offset);          m_data->trace(ev.m_data.get(), axis_offset);
1914          return ev;          return ev;
1915       }       }
1916       if (getDataPointRank()==4) {       if (getDataPointRank()==4) {
1917          DataArrayView::ShapeType ev_shape;          DataTypes::ShapeType ev_shape;
1918          if (axis_offset==0) {          if (axis_offset==0) {
1919            ev_shape.push_back(s[2]);            ev_shape.push_back(s[2]);
1920            ev_shape.push_back(s[3]);            ev_shape.push_back(s[3]);
# Line 1432  Data::matrixtrace(int axis_offset) const Line 1929  Data::matrixtrace(int axis_offset) const
1929      }      }
1930          Data ev(0.,ev_shape,getFunctionSpace());          Data ev(0.,ev_shape,getFunctionSpace());
1931          ev.typeMatchRight(*this);          ev.typeMatchRight(*this);
1932      m_data->matrixtrace(ev.m_data.get(), axis_offset);      m_data->trace(ev.m_data.get(), axis_offset);
1933          return ev;          return ev;
1934       }       }
1935       else {       else {
1936          throw DataException("Error - Data::matrixtrace can only be calculated for rank 2, 3 or 4 object.");          throw DataException("Error - Data::trace can only be calculated for rank 2, 3 or 4 object.");
1937       }       }
1938  }  }
1939    
1940  Data  Data
1941  Data::transpose(int axis_offset) const  Data::transpose(int axis_offset) const
1942  {  {    
1943  #if defined DOPROF       MAKELAZYOPOFF(TRANS,axis_offset)
1944       profData->reduction2++;       DataTypes::ShapeType s=getDataPointShape();
1945  #endif       DataTypes::ShapeType ev_shape;
      DataArrayView::ShapeType s=getDataPointShape();  
      DataArrayView::ShapeType ev_shape;  
1946       // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]       // Here's the equivalent of python s_out=s[axis_offset:]+s[:axis_offset]
1947       // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)       // which goes thru all shape vector elements starting with axis_offset (at index=rank wrap around to 0)
1948       int rank=getDataPointRank();       int rank=getDataPointRank();
# Line 1455  Data::transpose(int axis_offset) const Line 1950  Data::transpose(int axis_offset) const
1950          throw DataException("Error - Data::transpose must have 0 <= axis_offset <= rank=" + rank);          throw DataException("Error - Data::transpose must have 0 <= axis_offset <= rank=" + rank);
1951       }       }
1952       for (int i=0; i<rank; i++) {       for (int i=0; i<rank; i++) {
1953    
1954         int index = (axis_offset+i)%rank;         int index = (axis_offset+i)%rank;
1955         ev_shape.push_back(s[index]); // Append to new shape         ev_shape.push_back(s[index]); // Append to new shape
1956       }       }
# Line 1467  Data::transpose(int axis_offset) const Line 1963  Data::transpose(int axis_offset) const
1963  Data  Data
1964  Data::eigenvalues() const  Data::eigenvalues() const
1965  {  {
1966       #if defined DOPROF       if (isLazy())
1967          profData->unary++;       {
1968       #endif      Data temp(*this);   // to get around the fact that you can't resolve a const Data
1969        temp.resolve();
1970        return temp.eigenvalues();
1971         }
1972       // check input       // check input
1973       DataArrayView::ShapeType s=getDataPointShape();       DataTypes::ShapeType s=getDataPointShape();
1974       if (getDataPointRank()!=2)       if (getDataPointRank()!=2)
1975          throw DataException("Error - Data::eigenvalues can only be calculated for rank 2 object.");          throw DataException("Error - Data::eigenvalues can only be calculated for rank 2 object.");
1976       if(s[0] != s[1])       if(s[0] != s[1])
1977          throw DataException("Error - Data::eigenvalues can only be calculated for object with equal first and second dimension.");          throw DataException("Error - Data::eigenvalues can only be calculated for object with equal first and second dimension.");
1978       // create return       // create return
1979       DataArrayView::ShapeType ev_shape(1,s[0]);       DataTypes::ShapeType ev_shape(1,s[0]);
1980       Data ev(0.,ev_shape,getFunctionSpace());       Data ev(0.,ev_shape,getFunctionSpace());
1981       ev.typeMatchRight(*this);       ev.typeMatchRight(*this);
1982       m_data->eigenvalues(ev.m_data.get());       m_data->eigenvalues(ev.m_data.get());
# Line 1487  Data::eigenvalues() const Line 1986  Data::eigenvalues() const
1986  const boost::python::tuple  const boost::python::tuple
1987  Data::eigenvalues_and_eigenvectors(const double tol) const  Data::eigenvalues_and_eigenvectors(const double tol) const
1988  {  {
1989       #if defined DOPROF       if (isLazy())
1990          profData->unary++;       {
1991       #endif      Data temp(*this);   // to get around the fact that you can't resolve a const Data
1992       DataArrayView::ShapeType s=getDataPointShape();      temp.resolve();
1993       if (getDataPointRank()!=2)      return temp.eigenvalues_and_eigenvectors(tol);
1994         }
1995         DataTypes::ShapeType s=getDataPointShape();
1996         if (getDataPointRank()!=2)
1997          throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for rank 2 object.");          throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for rank 2 object.");
1998       if(s[0] != s[1])       if(s[0] != s[1])
1999          throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for object with equal first and second dimension.");          throw DataException("Error - Data::eigenvalues and eigenvectors can only be calculated for object with equal first and second dimension.");
2000       // create return       // create return
2001       DataArrayView::ShapeType ev_shape(1,s[0]);       DataTypes::ShapeType ev_shape(1,s[0]);
2002       Data ev(0.,ev_shape,getFunctionSpace());       Data ev(0.,ev_shape,getFunctionSpace());
2003       ev.typeMatchRight(*this);       ev.typeMatchRight(*this);
2004       DataArrayView::ShapeType V_shape(2,s[0]);       DataTypes::ShapeType V_shape(2,s[0]);
2005       Data V(0.,V_shape,getFunctionSpace());       Data V(0.,V_shape,getFunctionSpace());
2006       V.typeMatchRight(*this);       V.typeMatchRight(*this);
2007       m_data->eigenvalues_and_eigenvectors(ev.m_data.get(),V.m_data.get(),tol);       m_data->eigenvalues_and_eigenvectors(ev.m_data.get(),V.m_data.get(),tol);
# Line 1507  Data::eigenvalues_and_eigenvectors(const Line 2009  Data::eigenvalues_and_eigenvectors(const
2009  }  }
2010    
2011  const boost::python::tuple  const boost::python::tuple
2012  Data::mindp() const  Data::minGlobalDataPoint() const
2013  {  {
2014    // NB: calc_mindp had to be split off from mindp as boost::make_tuple causes an    // NB: calc_minGlobalDataPoint( had to be split off from minGlobalDataPoint( as boost::make_tuple causes an
2015    // abort (for unknown reasons) if there are openmp directives with it in the    // abort (for unknown reasons) if there are openmp directives with it in the
2016    // surrounding function    // surrounding function
2017    
   int SampleNo;  
2018    int DataPointNo;    int DataPointNo;
2019      int ProcNo;    int ProcNo;
2020    calc_mindp(ProcNo,SampleNo,DataPointNo);    calc_minGlobalDataPoint(ProcNo,DataPointNo);
2021    return make_tuple(ProcNo,SampleNo,DataPointNo);    return make_tuple(ProcNo,DataPointNo);
2022  }  }
2023    
2024  void  void
2025  Data::calc_mindp(   int& ProcNo,  Data::calc_minGlobalDataPoint(int& ProcNo,
2026                  int& SampleNo,                          int& DataPointNo) const
         int& DataPointNo) const  
2027  {  {
2028      if (isLazy())
2029      {
2030        Data temp(*this);   // to get around the fact that you can't resolve a const Data
2031        temp.resolve();
2032        return temp.calc_minGlobalDataPoint(ProcNo,DataPointNo);
2033      }
2034    int i,j;    int i,j;
2035    int lowi=0,lowj=0;    int lowi=0,lowj=0;
2036    double min=numeric_limits<double>::max();    double min=numeric_limits<double>::max();
2037    
2038    Data temp=minval();    Data temp=minval_nonlazy();   // need to do this to prevent autolazy from reintroducing laziness
2039    
2040    int numSamples=temp.getNumSamples();    int numSamples=temp.getNumSamples();
2041    int numDPPSample=temp.getNumDataPointsPerSample();    int numDPPSample=temp.getNumDataPointsPerSample();
2042    
2043    double next,local_min;    double next,local_min;
2044    int local_lowi,local_lowj;    int local_lowi=0,local_lowj=0;    
2045    
2046    #pragma omp parallel private(next,local_min,local_lowi,local_lowj)    #pragma omp parallel firstprivate(local_lowi,local_lowj) private(next,local_min)
2047    {    {
2048      local_min=min;      local_min=min;
2049      #pragma omp for private(i,j) schedule(static)      #pragma omp for private(i,j) schedule(static)
2050      for (i=0; i<numSamples; i++) {      for (i=0; i<numSamples; i++) {
2051        for (j=0; j<numDPPSample; j++) {        for (j=0; j<numDPPSample; j++) {
2052          next=temp.getDataPoint(i,j)();          next=temp.getDataAtOffsetRO(temp.getDataOffset(i,j));
2053          if (next<local_min) {          if (next<local_min) {
2054            local_min=next;            local_min=next;
2055            local_lowi=i;            local_lowi=i;
# Line 1552  Data::calc_mindp(  int& ProcNo, Line 2058  Data::calc_mindp(  int& ProcNo,
2058        }        }
2059      }      }
2060      #pragma omp critical      #pragma omp critical
2061      if (local_min<min) {      if (local_min<min) {    // If we found a smaller value than our sentinel
2062        min=local_min;        min=local_min;
2063        lowi=local_lowi;        lowi=local_lowi;
2064        lowj=local_lowj;        lowj=local_lowj;
# Line 1560  Data::calc_mindp(  int& ProcNo, Line 2066  Data::calc_mindp(  int& ProcNo,
2066    }    }
2067    
2068  #ifdef PASO_MPI  #ifdef PASO_MPI
2069      // determine the processor on which the minimum occurs    // determine the processor on which the minimum occurs
2070      next = temp.getDataPoint(lowi,lowj)();    next = temp.getDataPointRO(lowi,lowj);
2071      int lowProc = 0;    int lowProc = 0;
2072      double *globalMins = new double[get_MPISize()+1];    double *globalMins = new double[get_MPISize()+1];
2073      int error = MPI_Gather ( &next, 1, MPI_DOUBLE, globalMins, 1, MPI_DOUBLE, 0, get_MPIComm() );    int error;
2074          error = MPI_Gather ( &next, 1, MPI_DOUBLE, globalMins, 1, MPI_DOUBLE, 0, get_MPIComm() );
2075      if( get_MPIRank()==0 ){  
2076          next = globalMins[lowProc];    if( get_MPIRank()==0 ){
2077          for( i=1; i<get_MPISize(); i++ )      next = globalMins[lowProc];
2078              if( next>globalMins[i] ){      for( i=1; i<get_MPISize(); i++ )
2079                  lowProc = i;          if( next>globalMins[i] ){
2080                  next = globalMins[i];              lowProc = i;
2081              }              next = globalMins[i];
2082            }
2083      }
2084      MPI_Bcast( &lowProc, 1, MPI_INT, 0, get_MPIComm() );
2085      DataPointNo = lowj + lowi * numDPPSample;
2086      MPI_Bcast(&DataPointNo, 1, MPI_INT, lowProc, get_MPIComm() );
2087      delete [] globalMins;
2088      ProcNo = lowProc;
2089    #else
2090      ProcNo = 0;
2091      DataPointNo = lowj + lowi * numDPPSample;
2092    #endif
2093    }
2094    
2095    
2096    const boost::python::tuple
2097    Data::maxGlobalDataPoint() const
2098    {
2099      int DataPointNo;
2100      int ProcNo;
2101      calc_maxGlobalDataPoint(ProcNo,DataPointNo);
2102      return make_tuple(ProcNo,DataPointNo);
2103    }
2104    
2105    void
2106    Data::calc_maxGlobalDataPoint(int& ProcNo,
2107                            int& DataPointNo) const
2108    {
2109      if (isLazy())
2110      {
2111        Data temp(*this);   // to get around the fact that you can't resolve a const Data
2112        temp.resolve();
2113        return temp.calc_maxGlobalDataPoint(ProcNo,DataPointNo);
2114      }
2115      int i,j;
2116      int highi=0,highj=0;
2117    //-------------
2118      double max= -numeric_limits<double>::max();
2119    
2120      Data temp=maxval_nonlazy();   // need to do this to prevent autolazy from reintroducing laziness
2121    
2122      int numSamples=temp.getNumSamples();
2123      int numDPPSample=temp.getNumDataPointsPerSample();
2124    
2125      double next,local_max;
2126      int local_highi=0,local_highj=0;  
2127    
2128      #pragma omp parallel firstprivate(local_highi,local_highj) private(next,local_max)
2129      {
2130        local_max=max;
2131        #pragma omp for private(i,j) schedule(static)
2132        for (i=0; i<numSamples; i++) {
2133          for (j=0; j<numDPPSample; j++) {
2134            next=temp.getDataAtOffsetRO(temp.getDataOffset(i,j));
2135            if (next>local_max) {
2136              local_max=next;
2137              local_highi=i;
2138              local_highj=j;
2139            }
2140          }
2141        }
2142        #pragma omp critical
2143        if (local_max>max) {    // If we found a larger value than our sentinel
2144          max=local_max;
2145          highi=local_highi;
2146          highj=local_highj;
2147        }
2148      }
2149    #ifdef PASO_MPI
2150      // determine the processor on which the maximum occurs
2151      next = temp.getDataPointRO(highi,highj);
2152      int highProc = 0;
2153      double *globalMaxs = new double[get_MPISize()+1];
2154      int error;
2155      error = MPI_Gather ( &next, 1, MPI_DOUBLE, globalMaxs, 1, MPI_DOUBLE, 0, get_MPIComm() );
2156      if( get_MPIRank()==0 ){
2157        next = globalMaxs[highProc];
2158        for( i=1; i<get_MPISize(); i++ )
2159        {
2160        if( next<globalMaxs[i] )
2161        {
2162            highProc = i;
2163            next = globalMaxs[i];
2164      }      }
2165      MPI_Bcast( &lowProc, 1, MPI_DOUBLE, 0, get_MPIComm() );      }
2166      }
2167      MPI_Bcast( &highProc, 1, MPI_INT, 0, get_MPIComm() );
2168      DataPointNo = highj + highi * numDPPSample;  
2169      MPI_Bcast(&DataPointNo, 1, MPI_INT, highProc, get_MPIComm() );
2170    
2171      delete [] globalMins;    delete [] globalMaxs;
2172      ProcNo = lowProc;    ProcNo = highProc;
2173  #else  #else
2174      ProcNo = 0;    ProcNo = 0;
2175      DataPointNo = highj + highi * numDPPSample;
2176  #endif  #endif
   SampleNo = lowi;  
   DataPointNo = lowj;  
2177  }  }
2178    
2179  void  void
2180  Data::saveDX(std::string fileName) const  Data::saveDX(std::string fileName) const
2181  {  {
2182      if (isEmpty())
2183      {
2184        throw DataException("Error - Operations not permitted on instances of DataEmpty.");
2185      }
2186      if (isLazy())
2187      {
2188         Data temp(*this);  // to get around the fact that you can't resolve a const Data
2189         temp.resolve();
2190         temp.saveDX(fileName);
2191         return;
2192      }
2193    boost::python::dict args;    boost::python::dict args;
2194    args["data"]=boost::python::object(this);    args["data"]=boost::python::object(this);
2195    getDomain().saveDX(fileName,args);    getDomain()->saveDX(fileName,args);
2196    return;    return;
2197  }  }
2198    
2199  void  void
2200  Data::saveVTK(std::string fileName) const  Data::saveVTK(std::string fileName) const
2201  {  {
2202      if (isEmpty())
2203      {
2204        throw DataException("Error - Operations not permitted on instances of DataEmpty.");
2205      }
2206      if (isLazy())
2207      {
2208         Data temp(*this);  // to get around the fact that you can't resolve a const Data
2209         temp.resolve();
2210         temp.saveVTK(fileName);
2211         return;
2212      }
2213    boost::python::dict args;    boost::python::dict args;
2214    args["data"]=boost::python::object(this);    args["data"]=boost::python::object(this);
2215    getDomain().saveVTK(fileName,args);    getDomain()->saveVTK(fileName,args,"","");
2216    return;    return;
2217  }  }
2218    
2219    
2220    
2221  Data&  Data&
2222  Data::operator+=(const Data& right)  Data::operator+=(const Data& right)
2223  {  {
2224    if (isProtected()) {    if (isProtected()) {
2225          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2226    }    }
2227  #if defined DOPROF    MAKELAZYBINSELF(right,ADD)    // for lazy + is equivalent to +=
2228    profData->binary++;    exclusiveWrite();         // Since Lazy data does not modify its leaves we only need to worry here
 #endif  
2229    binaryOp(right,plus<double>());    binaryOp(right,plus<double>());
2230    return (*this);    return (*this);
2231  }  }
# Line 1622  Data::operator+=(const boost::python::ob Line 2236  Data::operator+=(const boost::python::ob
2236    if (isProtected()) {    if (isProtected()) {
2237          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2238    }    }
2239  #if defined DOPROF    Data tmp(right,getFunctionSpace(),false);
2240    profData->binary++;    (*this)+=tmp;
2241  #endif    return *this;
2242    binaryOp(right,plus<double>());  }
2243    
2244    // Hmmm, operator= makes a deep copy but the copy constructor does not?
2245    Data&
2246    Data::operator=(const Data& other)
2247    {
2248      m_protected=false;        // since any changes should be caught by exclusiveWrite();
2249    //   m_data=other.m_data;
2250      set_m_data(other.m_data);
2251    return (*this);    return (*this);
2252  }  }
2253    
# Line 1635  Data::operator-=(const Data& right) Line 2257  Data::operator-=(const Data& right)
2257    if (isProtected()) {    if (isProtected()) {
2258          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2259    }    }
2260  #if defined DOPROF    MAKELAZYBINSELF(right,SUB)
2261    profData->binary++;    exclusiveWrite();
 #endif  
2262    binaryOp(right,minus<double>());    binaryOp(right,minus<double>());
2263    return (*this);    return (*this);
2264  }  }
# Line 1648  Data::operator-=(const boost::python::ob Line 2269  Data::operator-=(const boost::python::ob
2269    if (isProtected()) {    if (isProtected()) {
2270          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2271    }    }
2272  #if defined DOPROF    Data tmp(right,getFunctionSpace(),false);
2273    profData->binary++;    (*this)-=tmp;
 #endif  
   binaryOp(right,minus<double>());  
2274    return (*this);    return (*this);
2275  }  }
2276    
# Line 1661  Data::operator*=(const Data& right) Line 2280  Data::operator*=(const Data& right)
2280    if (isProtected()) {    if (isProtected()) {
2281          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2282    }    }
2283  #if defined DOPROF    MAKELAZYBINSELF(right,MUL)
2284    profData->binary++;    exclusiveWrite();
 #endif  
2285    binaryOp(right,multiplies<double>());    binaryOp(right,multiplies<double>());
2286    return (*this);    return (*this);
2287  }  }
2288    
2289  Data&  Data&
2290  Data::operator*=(const boost::python::object& right)  Data::operator*=(const boost::python::object& right)
2291  {  {  
2292    if (isProtected()) {    if (isProtected()) {
2293          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2294    }    }
2295  #if defined DOPROF    Data tmp(right,getFunctionSpace(),false);
2296    profData->binary++;    (*this)*=tmp;
 #endif  
   binaryOp(right,multiplies<double>());  
2297    return (*this);    return (*this);
2298  }  }
2299    
# Line 1687  Data::operator/=(const Data& right) Line 2303  Data::operator/=(const Data& right)
2303    if (isProtected()) {    if (isProtected()) {
2304          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2305    }    }
2306  #if defined DOPROF    MAKELAZYBINSELF(right,DIV)
2307    profData->binary++;    exclusiveWrite();
 #endif  
2308    binaryOp(right,divides<double>());    binaryOp(right,divides<double>());
2309    return (*this);    return (*this);
2310  }  }
# Line 1700  Data::operator/=(const boost::python::ob Line 2315  Data::operator/=(const boost::python::ob
2315    if (isProtected()) {    if (isProtected()) {
2316          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2317    }    }
2318  #if defined DOPROF    Data tmp(right,getFunctionSpace(),false);
2319    profData->binary++;    (*this)/=tmp;
 #endif  
   binaryOp(right,divides<double>());  
2320    return (*this);    return (*this);
2321  }  }
2322    
2323  Data  Data
2324  Data::rpowO(const boost::python::object& left) const  Data::rpowO(const boost::python::object& left) const
2325  {  {
   if (isProtected()) {  
         throw DataException("Error - attempt to update protected Data object.");  
   }  
 #if defined DOPROF  
   profData->binary++;  
 #endif  
2326    Data left_d(left,*this);    Data left_d(left,*this);
2327    return left_d.powD(*this);    return left_d.powD(*this);
2328  }  }
# Line 1723  Data::rpowO(const boost::python::object& Line 2330  Data::rpowO(const boost::python::object&
2330  Data  Data
2331  Data::powO(const boost::python::object& right) const  Data::powO(const boost::python::object& right) const
2332  {  {
2333  #if defined DOPROF    Data tmp(right,getFunctionSpace(),false);
2334    profData->binary++;    return powD(tmp);
 #endif  
   Data result;  
   result.copy(*this);  
   result.binaryOp(right,(Data::BinaryDFunPtr)::pow);  
   return result;  
2335  }  }
2336    
2337  Data  Data
2338  Data::powD(const Data& right) const  Data::powD(const Data& right) const
2339  {  {
2340  #if defined DOPROF    MAKELAZYBIN(right,POW)
2341    profData->binary++;    return C_TensorBinaryOperation<double (*)(double, double)>(*this, right, ::pow);
 #endif  
   Data result;  
   result.copy(*this);  
   result.binaryOp(right,(Data::BinaryDFunPtr)::pow);  
   return result;  
2342  }  }
2343    
   
2344  //  //
2345  // NOTE: It is essential to specify the namespace this operator belongs to  // NOTE: It is essential to specify the namespace this operator belongs to
2346  Data  Data
2347  escript::operator+(const Data& left, const Data& right)  escript::operator+(const Data& left, const Data& right)
2348  {  {
2349    Data result;    MAKELAZYBIN2(left,right,ADD)
2350    //    return C_TensorBinaryOperation(left, right, plus<double>());
   // perform a deep copy  
   result.copy(left);  
   result+=right;  
   return result;  
2351  }  }
2352    
2353  //  //
# Line 1763  escript::operator+(const Data& left, con Line 2355  escript::operator+(const Data& left, con
2355  Data  Data
2356  escript::operator-(const Data& left, const Data& right)  escript::operator-(const Data& left, const Data& right)
2357  {  {
2358    Data result;    MAKELAZYBIN2(left,right,SUB)
2359    //    return C_TensorBinaryOperation(left, right, minus<double>());
   // perform a deep copy  
   result.copy(left);  
   result-=right;  
   return result;  
2360  }  }
2361    
2362  //  //
# Line 1776  escript::operator-(const Data& left, con Line 2364  escript::operator-(const Data& left, con
2364  Data  Data
2365  escript::operator*(const Data& left, const Data& right)  escript::operator*(const Data& left, const Data& right)
2366  {  {
2367    Data result;    MAKELAZYBIN2(left,right,MUL)
2368    //    return C_TensorBinaryOperation(left, right, multiplies<double>());
   // perform a deep copy  
   result.copy(left);  
   result*=right;  
   return result;  
2369  }  }
2370    
2371  //  //
# Line 1789  escript::operator*(const Data& left, con Line 2373  escript::operator*(const Data& left, con
2373  Data  Data
2374  escript::operator/(const Data& left, const Data& right)  escript::operator/(const Data& left, const Data& right)
2375  {  {
2376    Data result;    MAKELAZYBIN2(left,right,DIV)
2377    //    return C_TensorBinaryOperation(left, right, divides<double>());
   // perform a deep copy  
   result.copy(left);  
   result/=right;  
   return result;  
2378  }  }
2379    
2380  //  //
# Line 1802  escript::operator/(const Data& left, con Line 2382  escript::operator/(const Data& left, con
2382  Data  Data
2383  escript::operator+(const Data& left, const boost::python::object& right)  escript::operator+(const Data& left, const boost::python::object& right)
2384  {  {
2385    //    Data tmp(right,left.getFunctionSpace(),false);
2386    // Convert to DataArray format if possible    MAKELAZYBIN2(left,tmp,ADD)
2387    DataArray temp(right);    return left+tmp;
   Data result;  
   //  
   // perform a deep copy  
   result.copy(left);  
   result+=right;  
   return result;  
2388  }  }
2389    
2390  //  //
# Line 1818  escript::operator+(const Data& left, con Line 2392  escript::operator+(const Data& left, con
2392  Data  Data
2393  escript::operator-(const Data& left, const boost::python::object& right)  escript::operator-(const Data& left, const boost::python::object& right)
2394  {  {
2395    //    Data tmp(right,left.getFunctionSpace(),false);
2396    // Convert to DataArray format if possible    MAKELAZYBIN2(left,tmp,SUB)
2397    DataArray temp(right);    return left-tmp;
   Data result;  
   //  
   // perform a deep copy  
   result.copy(left);  
   result-=right;  
   return result;  
2398  }  }
2399    
2400  //  //
# Line 1834  escript::operator-(const Data& left, con Line 2402  escript::operator-(const Data& left, con
2402  Data  Data
2403  escript::operator*(const Data& left, const boost::python::object& right)  escript::operator*(const Data& left, const boost::python::object& right)
2404  {  {
2405    //    Data tmp(right,left.getFunctionSpace(),false);
2406    // Convert to DataArray format if possible    MAKELAZYBIN2(left,tmp,MUL)
2407    DataArray temp(right);    return left*tmp;
   Data result;  
   //  
   // perform a deep copy  
   result.copy(left);  
   result*=right;  
   return result;  
2408  }  }
2409    
2410  //  //
# Line 1850  escript::operator*(const Data& left, con Line 2412  escript::operator*(const Data& left, con
2412  Data  Data
2413  escript::operator/(const Data& left, const boost::python::object& right)  escript::operator/(const Data& left, const boost::python::object& right)
2414  {  {
2415    //    Data tmp(right,left.getFunctionSpace(),false);
2416    // Convert to DataArray format if possible    MAKELAZYBIN2(left,tmp,DIV)
2417    DataArray temp(right);    return left/tmp;
   Data result;  
   //  
   // perform a deep copy  
   result.copy(left);  
   result/=right;  
   return result;  
2418  }  }
2419    
2420  //  //
# Line 1866  escript::operator/(const Data& left, con Line 2422  escript::operator/(const Data& left, con
2422  Data  Data
2423  escript::operator+(const boost::python::object& left, const Data& right)  escript::operator+(const boost::python::object& left, const Data& right)
2424  {  {
2425    //    Data tmp(left,right.getFunctionSpace(),false);
2426    // Construct the result using the given value and the other parameters    MAKELAZYBIN2(tmp,right,ADD)
2427    // from right    return tmp+right;
   Data result(left,right);  
   result+=right;  
   return result;  
2428  }  }
2429    
2430  //  //
# Line 1879  escript::operator+(const boost::python:: Line 2432  escript::operator+(const boost::python::
2432  Data  Data
2433  escript::operator-(const boost::python::object& left, const Data& right)  escript::operator-(const boost::python::object& left, const Data& right)
2434  {  {
2435    //    Data tmp(left,right.getFunctionSpace(),false);
2436    // Construct the result using the given value and the other parameters    MAKELAZYBIN2(tmp,right,SUB)
2437    // from right    return tmp-right;
   Data result(left,right);  
   result-=right;  
   return result;  
2438  }  }
2439    
2440  //  //
# Line 1892  escript::operator-(const boost::python:: Line 2442  escript::operator-(const boost::python::
2442  Data  Data
2443  escript::operator*(const boost::python::object& left, const Data& right)  escript::operator*(const boost::python::object& left, const Data& right)
2444  {  {
2445    //    Data tmp(left,right.getFunctionSpace(),false);
2446    // Construct the result using the given value and the other parameters    MAKELAZYBIN2(tmp,right,MUL)
2447    // from right    return tmp*right;
   Data result(left,right);  
   result*=right;  
   return result;  
2448  }  }
2449    
2450  //  //
# Line 1905  escript::operator*(const boost::python:: Line 2452  escript::operator*(const boost::python::
2452  Data  Data
2453  escript::operator/(const boost::python::object& left, const Data& right)  escript::operator/(const boost::python::object& left, const Data& right)
2454  {  {
2455    //    Data tmp(left,right.getFunctionSpace(),false);
2456    // Construct the result using the given value and the other parameters    MAKELAZYBIN2(tmp,right,DIV)
2457    // from right    return tmp/right;
   Data result(left,right);  
   result/=right;  
   return result;  
2458  }  }
2459    
 //  
 //bool escript::operator==(const Data& left, const Data& right)  
 //{  
 //  /*  
 //  NB: this operator does very little at this point, and isn't to  
 //  be relied on. Requires further implementation.  
 //  */  
 //  
 //  bool ret;  
 //  
 //  if (left.isEmpty()) {  
 //    if(!right.isEmpty()) {  
 //      ret = false;  
 //    } else {  
 //      ret = true;  
 //    }  
 //  }  
 //  
 //  if (left.isConstant()) {  
 //    if(!right.isConstant()) {  
 //      ret = false;  
 //    } else {  
 //      ret = true;  
 //    }  
 // }  
 //  
 //  if (left.isTagged()) {  
 //   if(!right.isTagged()) {  
 //      ret = false;  
 //    } else {  
 //      ret = true;  
 //    }  
 //  }  
 //  
 //  if (left.isExpanded()) {  
 //    if(!right.isExpanded()) {  
 //      ret = false;  
 //    } else {  
 //      ret = true;  
 //    }  
 //  }  
 //  
 //  return ret;  
 //}  
2460    
2461  /* TODO */  /* TODO */
2462  /* global reduction */  /* global reduction */
2463  Data  Data
2464  Data::getItem(const boost::python::object& key) const  Data::getItem(const boost::python::object& key) const
2465  {  {
   const DataArrayView& view=getPointDataView();  
2466    
2467    DataArrayView::RegionType slice_region=view.getSliceRegion(key);    DataTypes::RegionType slice_region=DataTypes::getSliceRegion(getDataPointShape(),key);
2468    
2469    if (slice_region.size()!=view.getRank()) {    if (slice_region.size()!=getDataPointRank()) {
2470      throw DataException("Error - slice size does not match Data rank.");      throw DataException("Error - slice size does not match Data rank.");
2471    }    }
2472    
# Line 1977  Data::getItem(const boost::python::objec Line 2476  Data::getItem(const boost::python::objec
2476  /* TODO */  /* TODO */
2477  /* global reduction */  /* global reduction */
2478  Data  Data
2479  Data::getSlice(const DataArrayView::RegionType& region) const  Data::getSlice(const DataTypes::RegionType& region) const
2480  {  {
 #if defined DOPROF  
   profData->slicing++;  
 #endif  
2481    return Data(*this,region);    return Data(*this,region);
2482  }  }
2483    
# Line 1995  Data::setItemO(const boost::python::obje Line 2491  Data::setItemO(const boost::python::obje
2491    setItemD(key,tempData);    setItemD(key,tempData);
2492  }  }
2493    
 /* TODO */  
 /* global reduction */  
2494  void  void
2495  Data::setItemD(const boost::python::object& key,  Data::setItemD(const boost::python::object& key,
2496                 const Data& value)                 const Data& value)
2497  {  {
2498    const DataArrayView& view=getPointDataView();    DataTypes::RegionType slice_region=DataTypes::getSliceRegion(getDataPointShape(),key);
2499      if (slice_region.size()!=getDataPointRank()) {
   DataArrayView::RegionType slice_region=view.getSliceRegion(key);  
   if (slice_region.size()!=view.getRank()) {  
2500      throw DataException("Error - slice size does not match Data rank.");      throw DataException("Error - slice size does not match Data rank.");
2501    }    }
2502      exclusiveWrite();
2503    if (getFunctionSpace()!=value.getFunctionSpace()) {    if (getFunctionSpace()!=value.getFunctionSpace()) {
2504       setSlice(Data(value,getFunctionSpace()),slice_region);       setSlice(Data(value,getFunctionSpace()),slice_region);
2505    } else {    } else {
# Line 2014  Data::setItemD(const boost::python::obje Line 2507  Data::setItemD(const boost::python::obje
2507    }    }
2508  }  }
2509    
 /* TODO */  
 /* global reduction */  
2510  void  void
2511  Data::setSlice(const Data& value,  Data::setSlice(const Data& value,
2512                 const DataArrayView::RegionType& region)                 const DataTypes::RegionType& region)
2513  {  {
2514    if (isProtected()) {    if (isProtected()) {
2515          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2516    }    }
2517  #if defined DOPROF    forceResolve();
2518    profData->slicing++;    exclusiveWrite();     // In case someone finds a way to call this without going through setItemD
 #endif  
2519    Data tempValue(value);    Data tempValue(value);
2520    typeMatchLeft(tempValue);    typeMatchLeft(tempValue);
2521    typeMatchRight(tempValue);    typeMatchRight(tempValue);
2522    m_data->setSlice(tempValue.m_data.get(),region);    getReady()->setSlice(tempValue.m_data.get(),region);
2523  }  }
2524    
2525  void  void
2526  Data::typeMatchLeft(Data& right) const  Data::typeMatchLeft(Data& right) const
2527  {  {
2528      if (right.isLazy() && !isLazy())
2529      {
2530        right.resolve();
2531      }
2532    if (isExpanded()){    if (isExpanded()){
2533      right.expand();      right.expand();
2534    } else if (isTagged()) {    } else if (isTagged()) {
# Line 2047  Data::typeMatchLeft(Data& right) const Line 2541  Data::typeMatchLeft(Data& right) const
2541  void  void
2542  Data::typeMatchRight(const Data& right)  Data::typeMatchRight(const Data& right)
2543  {  {
2544      if (isLazy() && !right.isLazy())
2545      {
2546        resolve();
2547      }
2548    if (isTagged()) {    if (isTagged()) {
2549      if (right.isExpanded()) {      if (right.isExpanded()) {
2550        expand();        expand();
# Line 2060  Data::typeMatchRight(const Data& right) Line 2558  Data::typeMatchRight(const Data& right)
2558    }    }
2559  }  }
2560    
2561  /* TODO */  // The normal TaggedValue adds the tag if it is not already present
2562  /* global reduction */  // This form does not. It throws instead.
2563    // This is because the names are maintained by the domain and cannot be added
2564    // without knowing the tag number to map it to.
2565    void
2566    Data::setTaggedValueByName(std::string name,
2567                               const boost::python::object& value)
2568    {
2569         if (getFunctionSpace().getDomain()->isValidTagName(name)) {
2570        forceResolve();
2571        exclusiveWrite();
2572            int tagKey=getFunctionSpace().getDomain()->getTag(name);
2573            setTaggedValue(tagKey,value);
2574         }
2575         else
2576         {                  // The
2577        throw DataException("Error - unknown tag in setTaggedValueByName.");
2578         }
2579    }
2580    
2581  void  void
2582  Data::setTaggedValue(int tagKey,  Data::setTaggedValue(int tagKey,
2583                       const boost::python::object& value)                       const boost::python::object& value)
# Line 2071  Data::setTaggedValue(int tagKey, Line 2587  Data::setTaggedValue(int tagKey,
2587    }    }
2588    //    //
2589    // Ensure underlying data object is of type DataTagged    // Ensure underlying data object is of type DataTagged
2590    tag();    forceResolve();
2591      exclusiveWrite();
2592      if (isConstant()) tag();
2593      WrappedArray w(value);
2594    
2595    if (!isTagged()) {    DataVector temp_data2;
2596      throw DataException("Error - DataTagged conversion failed!!");    temp_data2.copyFromArray(w,1);
   }  
2597    
2598    //    m_data->setTaggedValue(tagKey,w.getShape(), temp_data2);
   // Construct DataArray from boost::python::object input value  
   DataArray valueDataArray(value);  
   
   //  
   // Call DataAbstract::setTaggedValue  
   m_data->setTaggedValue(tagKey,valueDataArray.getView());  
2599  }  }
2600    
2601  /* TODO */  
 /* global reduction */  
2602  void  void
2603  Data::setTaggedValueFromCPP(int tagKey,  Data::setTaggedValueFromCPP(int tagKey,
2604                              const DataArrayView& value)                  const DataTypes::ShapeType& pointshape,
2605                                const DataTypes::ValueType& value,
2606                    int dataOffset)
2607  {  {
2608    if (isProtected()) {    if (isProtected()) {
2609          throw DataException("Error - attempt to update protected Data object.");          throw DataException("Error - attempt to update protected Data object.");
2610    }    }
2611    //    //
2612    // Ensure underlying data object is of type DataTagged    // Ensure underlying data object is of type DataTagged
2613    tag();    forceResolve();
2614      if (isConstant()) tag();
2615    if (!isTagged()) {    exclusiveWrite();
     throw DataException("Error - DataTagged conversion failed!!");  
   }  
                                                                                                                 
2616    //    //
2617    // Call DataAbstract::setTaggedValue    // Call DataAbstract::setTaggedValue
2618    m_data->setTaggedValue(tagKey,value);    m_data->setTaggedValue(tagKey,pointshape, value, dataOffset);
2619  }  }
2620    
 /* TODO */  
 /* global reduction */  
2621  int  int
2622  Data::getTagNumber(int dpno)  Data::getTagNumber(int dpno)
2623  {  {
2624    return m_data->getTagNumber(dpno);    if (isEmpty())
2625  }    {
2626        throw DataException("Error - operation not permitted on instances of DataEmpty.");
 /* TODO */  
 /* global reduction */  
 void  
 Data::setRefValue(int ref,  
                   const boost::python::numeric::array& value)  
 {  
   if (isProtected()) {  
         throw DataException("Error - attempt to update protected Data object.");  
2627    }    }
2628    //    return getFunctionSpace().getTagFromDataPointNo(dpno);
   // Construct DataArray from boost::python::object input value  
   DataArray valueDataArray(value);  
   
   //  
   // Call DataAbstract::setRefValue  
   m_data->setRefValue(ref,valueDataArray);  
2629  }  }
2630    
 /* TODO */  
 /* global reduction */  
 void  
 Data::getRefValue(int ref,  
                   boost::python::numeric::array& value)  
 {  
   //  
   // Construct DataArray for boost::python::object return value  
   DataArray valueDataArray(value);  
   
   //  
   // Load DataArray with values from data-points specified by ref  
   m_data->getRefValue(ref,valueDataArray);  
2631    
2632    //  ostream& escript::operator<<(ostream& o, const Data& data)
2633    // Load values from valueDataArray into return numarray  {
2634      o << data.toString();
2635      return o;
2636    }
2637    
2638    // extract the shape of the numarray  Data
2639    int rank = value.getrank();  escript::C_GeneralTensorProduct(Data& arg_0,
2640    DataArrayView::ShapeType shape;                       Data& arg_1,
2641    for (int i=0; i < rank; i++) {                       int axis_offset,
2642      shape.push_back(extract<int>(value.getshape()[i]));                       int transpose)
2643    {
2644      // General tensor product: res(SL x SR) = arg_0(SL x SM) * arg_1(SM x SR)
2645      // SM is the product of the last axis_offset entries in arg_0.getShape().
2646    
2647      // deal with any lazy data
2648    //   if (arg_0.isLazy()) {arg_0.resolve();}
2649    //   if (arg_1.isLazy()) {arg_1.resolve();}
2650      if (arg_0.isLazy() || arg_1.isLazy() || (AUTOLAZYON && (arg_0.isExpanded() || arg_1.isExpanded())))
2651      {
2652        DataLazy* c=new DataLazy(arg_0.borrowDataPtr(), arg_1.borrowDataPtr(), PROD, axis_offset,transpose);
2653        return Data(c);
2654    }    }
2655    
2656    // and load the numarray with the data from the DataArray    // Interpolate if necessary and find an appropriate function space
2657    DataArrayView valueView = valueDataArray.getView();    Data arg_0_Z, arg_1_Z;
2658      if (arg_0.getFunctionSpace()!=arg_1.getFunctionSpace()) {
2659    if (rank==0) {      if (arg_0.probeInterpolation(arg_1.getFunctionSpace())) {
2660        boost::python::numeric::array temp_numArray(valueView());        arg_0_Z = arg_0.interpolate(arg_1.getFunctionSpace());
2661        value = temp_numArray;        arg_1_Z = Data(arg_1);
2662    }      }
2663    if (rank==1) {      else if (arg_1.probeInterpolation(arg_0.getFunctionSpace())) {
2664      for (int i=0; i < shape[0]; i++) {        arg_1_Z=arg_1.interpolate(arg_0.getFunctionSpace());
2665        value[i] = valueView(i);        arg_0_Z =Data(arg_0);
2666      }      }
2667    }      else {
2668    if (rank==2) {        throw DataException("Error - C_GeneralTensorProduct: arguments have incompatible function spaces.");
     for (int i=0; i < shape[0]; i++) {  
       for (int j=0; j < shape[1]; j++) {  
         value[i][j] = valueView(i,j);  
       }  
2669      }      }
2670      } else {
2671          arg_0_Z = Data(arg_0);
2672          arg_1_Z = Data(arg_1);
2673    }    }
2674    if (rank==3) {    // Get rank and shape of inputs
2675      for (int i=0; i < shape[0]; i++) {    int rank0 = arg_0_Z.getDataPointRank();
2676        for (int j=0; j < shape[1]; j++) {    int rank1 = arg_1_Z.getDataPointRank();
2677          for (int k=0; k < shape[2]; k++) {    const DataTypes::ShapeType& shape0 = arg_0_Z.getDataPointShape();
2678            value[i][j][k] = valueView(i,j,k);    const DataTypes::ShapeType& shape1 = arg_1_Z.getDataPointShape();
2679          }  
2680        }    // Prepare for the loops of the product and verify compatibility of shapes
2681      int start0=0, start1=0;
2682      if (transpose == 0)       {}
2683      else if (transpose == 1)  { start0 = axis_offset; }
2684      else if (transpose == 2)  { start1 = rank1-axis_offset; }
2685      else              { throw DataException("C_GeneralTensorProduct: Error - transpose should be 0, 1 or 2"); }
2686    
2687    
2688      // Adjust the shapes for transpose
2689      DataTypes::ShapeType tmpShape0(rank0);    // pre-sizing the vectors rather
2690      DataTypes::ShapeType tmpShape1(rank1);    // than using push_back
2691      for (int i=0; i<rank0; i++)   { tmpShape0[i]=shape0[(i+start0)%rank0]; }
2692      for (int i=0; i<rank1; i++)   { tmpShape1[i]=shape1[(i+start1)%rank1]; }
2693    
2694    #if 0
2695      // For debugging: show shape after transpose
2696      char tmp[100];
2697      std::string shapeStr;
2698      shapeStr = "(";
2699      for (int i=0; i<rank0; i++)   { sprintf(tmp, "%d,", tmpShape0[i]); shapeStr += tmp; }
2700      shapeStr += ")";
2701      cout << "C_GeneralTensorProduct: Shape of arg0 is " << shapeStr << endl;
2702      shapeStr = "(";
2703      for (int i=0; i<rank1; i++)   { sprintf(tmp, "%d,", tmpShape1[i]); shapeStr += tmp; }
2704      shapeStr += ")";
2705      cout << "C_GeneralTensorProduct: Shape of arg1 is " << shapeStr << endl;
2706    #endif
2707    
2708      // Prepare for the loops of the product
2709      int SL=1, SM=1, SR=1;
2710      for (int i=0; i<rank0-axis_offset; i++)   {
2711        SL *= tmpShape0[i];
2712      }
2713      for (int i=rank0-axis_offset; i<rank0; i++)   {
2714        if (tmpShape0[i] != tmpShape1[i-(rank0-axis_offset)]) {
2715          throw DataException("C_GeneralTensorProduct: Error - incompatible shapes");
2716        }
2717        SM *= tmpShape0[i];
2718      }
2719      for (int i=axis_offset; i<rank1; i++)     {
2720        SR *= tmpShape1[i];
2721      }
2722    
2723      // Define the shape of the output (rank of shape is the sum of the loop ranges below)
2724      DataTypes::ShapeType shape2(rank0+rank1-2*axis_offset);  
2725      {         // block to limit the scope of out_index
2726         int out_index=0;
2727         for (int i=0; i<rank0-axis_offset; i++, ++out_index) { shape2[out_index]=tmpShape0[i]; } // First part of arg_0_Z
2728         for (int i=axis_offset; i<rank1; i++, ++out_index)   { shape2[out_index]=tmpShape1[i]; } // Last part of arg_1_Z
2729      }
2730    
2731      if (shape2.size()>ESCRIPT_MAX_DATA_RANK)
2732      {
2733         ostringstream os;
2734         os << "C_GeneralTensorProduct: Error - Attempt to create a rank " << shape2.size() << " object. The maximum rank is " << ESCRIPT_MAX_DATA_RANK << ".";
2735         throw DataException(os.str());
2736      }
2737    
2738      // Declare output Data object
2739      Data res;
2740    
2741      if      (arg_0_Z.isConstant()   && arg_1_Z.isConstant()) {
2742        res = Data(0.0, shape2, arg_1_Z.getFunctionSpace());    // DataConstant output
2743        const double *ptr_0 = &(arg_0_Z.getDataAtOffsetRO(0));
2744        const double *ptr_1 = &(arg_1_Z.getDataAtOffsetRO(0));
2745        double *ptr_2 = &(res.getDataAtOffsetRW(0));
2746        matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2747      }
2748      else if (arg_0_Z.isConstant()   && arg_1_Z.isTagged()) {
2749    
2750        // Prepare the DataConstant input
2751        DataConstant* tmp_0=dynamic_cast<DataConstant*>(arg_0_Z.borrowData());
2752        if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2753    
2754        // Borrow DataTagged input from Data object
2755        DataTagged* tmp_1=dynamic_cast<DataTagged*>(arg_1_Z.borrowData());
2756        if (tmp_1==0) { throw DataException("GTP_1 Programming error - casting to DataTagged."); }
2757    
2758        // Prepare a DataTagged output 2
2759        res = Data(0.0, shape2, arg_1_Z.getFunctionSpace());    // DataTagged output
2760        res.tag();
2761        DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2762        if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2763    
2764        // Prepare offset into DataConstant
2765        int offset_0 = tmp_0->getPointOffset(0,0);
2766        const double *ptr_0 = &(arg_0_Z.getDataAtOffsetRO(offset_0));
2767    
2768        const double *ptr_1 = &(tmp_1->getDefaultValueRO(0));
2769        double *ptr_2 = &(tmp_2->getDefaultValueRW(0));
2770    
2771        // Compute an MVP for the default
2772        matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2773        // Compute an MVP for each tag
2774        const DataTagged::DataMapType& lookup_1=tmp_1->getTagLookup();
2775        DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2776        for (i=lookup_1.begin();i!=lookup_1.end();i++) {
2777          tmp_2->addTag(i->first);
2778    
2779          const double *ptr_1 = &(tmp_1->getDataByTagRO(i->first,0));
2780          double *ptr_2 = &(tmp_2->getDataByTagRW(i->first,0));
2781        
2782          matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2783      }      }
2784    
2785    }    }
2786    if (rank==4) {    else if (arg_0_Z.isConstant()   && arg_1_Z.isExpanded()) {
2787      for (int i=0; i < shape[0]; i++) {  
2788        for (int j=0; j < shape[1]; j++) {      res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2789          for (int k=0; k < shape[2]; k++) {      DataConstant* tmp_0=dynamic_cast<DataConstant*>(arg_0_Z.borrowData());
2790            for (int l=0; l < shape[3]; l++) {      DataExpanded* tmp_1=dynamic_cast<DataExpanded*>(arg_1_Z.borrowData());
2791              value[i][j][k][l] = valueView(i,j,k,l);      DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2792            }      if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
2793          }      if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2794        if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2795        int sampleNo_1,dataPointNo_1;
2796        int numSamples_1 = arg_1_Z.getNumSamples();
2797        int numDataPointsPerSample_1 = arg_1_Z.getNumDataPointsPerSample();
2798        int offset_0 = tmp_0->getPointOffset(0,0);
2799        #pragma omp parallel for private(sampleNo_1,dataPointNo_1) schedule(static)
2800        for (sampleNo_1 = 0; sampleNo_1 < numSamples_1; sampleNo_1++) {
2801          for (dataPointNo_1 = 0; dataPointNo_1 < numDataPointsPerSample_1; dataPointNo_1++) {
2802            int offset_1 = tmp_1->getPointOffset(sampleNo_1,dataPointNo_1);
2803            int offset_2 = tmp_2->getPointOffset(sampleNo_1,dataPointNo_1);
2804            const double *ptr_0 = &(arg_0_Z.getDataAtOffsetRO(offset_0));
2805            const double *ptr_1 = &(arg_1_Z.getDataAtOffsetRO(offset_1));
2806            double *ptr_2 = &(res.getDataAtOffsetRW(offset_2));
2807            matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2808        }        }
2809      }      }
2810    
2811    }    }
2812      else if (arg_0_Z.isTagged()     && arg_1_Z.isConstant()) {
2813    
2814  }      // Borrow DataTagged input from Data object
2815        DataTagged* tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2816        if (tmp_0==0) { throw DataException("GTP_0 Programming error - casting to DataTagged."); }
2817    
2818  void      // Prepare the DataConstant input
2819  Data::archiveData(const std::string fileName)      DataConstant* tmp_1=dynamic_cast<DataConstant*>(arg_1_Z.borrowData());
2820  {      if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataConstant."); }
   cout << "Archiving Data object to: " << fileName << endl;  
2821    
2822    //      // Prepare a DataTagged output 2
2823    // Determine type of this Data object      res = Data(0.0, shape2, arg_0_Z.getFunctionSpace());    // DataTagged output
2824    int dataType = -1;      res.tag();
2825        DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2826        if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2827    
2828    if (isEmpty()) {      // Prepare offset into DataConstant
2829      dataType = 0;      int offset_1 = tmp_1->getPointOffset(0,0);
2830      cout << "\tdataType: DataEmpty" << endl;      const double *ptr_1 = &(arg_1_Z.getDataAtOffsetRO(offset_1));
2831    }      const double *ptr_0 = &(tmp_0->getDefaultValueRO(0));
2832    if (isConstant()) {      double *ptr_2 = &(tmp_2->getDefaultValueRW(0));
     dataType = 1;  
     cout << "\tdataType: DataConstant" << endl;  
   }  
   if (isTagged()) {  
     dataType = 2;  
     cout << "\tdataType: DataTagged" << endl;  
   }  
   if (isExpanded()) {  
     dataType = 3;  
     cout << "\tdataType: DataExpanded" << endl;  
   }  
2833    
2834    if (dataType == -1) {      // Compute an MVP for the default
2835      throw DataException("archiveData Error: undefined dataType");      matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2836    }      // Compute an MVP for each tag
2837        const DataTagged::DataMapType& lookup_0=tmp_0->getTagLookup();
2838        DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2839        for (i=lookup_0.begin();i!=lookup_0.end();i++) {
2840    
2841    //        tmp_2->addTag(i->first);
2842    // Collect data items common to all Data types        const double *ptr_0 = &(tmp_0->getDataByTagRO(i->first,0));
2843    int noSamples = getNumSamples();        double *ptr_2 = &(tmp_2->getDataByTagRW(i->first,0));
2844    int noDPPSample = getNumDataPointsPerSample();        matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
   int functionSpaceType = getFunctionSpace().getTypeCode();  
   int dataPointRank = getDataPointRank();  
   int dataPointSize = getDataPointSize();  
   int dataLength = getLength();  
   DataArrayView::ShapeType dataPointShape = getDataPointShape();  
   vector<int> referenceNumbers(noSamples);  
   for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {  
     referenceNumbers[sampleNo] = getFunctionSpace().getReferenceNoFromSampleNo(sampleNo);  
   }  
   vector<int> tagNumbers(noSamples);  
   if (isTagged()) {  
     for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {  
       tagNumbers[sampleNo] = getFunctionSpace().getTagFromSampleNo(sampleNo);  
2845      }      }
2846    
2847    }    }
2848      else if (arg_0_Z.isTagged()     && arg_1_Z.isTagged()) {
2849    
2850    cout << "\tnoSamples: " << noSamples << " noDPPSample: " << noDPPSample << endl;      // Borrow DataTagged input from Data object
2851    cout << "\tfunctionSpaceType: " << functionSpaceType << endl;      DataTagged* tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2852    cout << "\trank: " << dataPointRank << " size: " << dataPointSize << " length: " << dataLength << endl;      if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2853    
2854    //      // Borrow DataTagged input from Data object
2855    // Flatten Shape to an array of integers suitable for writing to file      DataTagged* tmp_1=dynamic_cast<DataTagged*>(arg_1_Z.borrowData());
2856    int flatShape[4] = {0,0,0,0};      if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
   cout << "\tshape: < ";  
   for (int dim=0; dim<dataPointRank; dim++) {  
     flatShape[dim] = dataPointShape[dim];  
     cout << dataPointShape[dim] << " ";  
   }  
   cout << ">" << endl;  
2857    
2858    //      // Prepare a DataTagged output 2
2859    // Open archive file      res = Data(0.0, shape2, arg_1_Z.getFunctionSpace());
2860    ofstream archiveFile;      res.tag();  // DataTagged output
2861    archiveFile.open(fileName.data(), ios::out);      DataTagged* tmp_2=dynamic_cast<DataTagged*>(res.borrowData());
2862        if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2863    
2864    if (!archiveFile.good()) {      const double *ptr_0 = &(tmp_0->getDefaultValueRO(0));
2865      throw DataException("archiveData Error: problem opening archive file");      const double *ptr_1 = &(tmp_1->getDefaultValueRO(0));
2866    }      double *ptr_2 = &(tmp_2->getDefaultValueRW(0));
2867    
2868    //      // Compute an MVP for the default
2869    // Write common data items to archive file      matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2870    archiveFile.write(reinterpret_cast<char *>(&dataType),sizeof(int));      // Merge the tags
2871    archiveFile.write(reinterpret_cast<char *>(&noSamples),sizeof(int));      DataTagged::DataMapType::const_iterator i; // i->first is a tag, i->second is an offset into memory
2872    archiveFile.write(reinterpret_cast<char *>(&noDPPSample),sizeof(int));      const DataTagged::DataMapType& lookup_0=tmp_0->getTagLookup();
2873    archiveFile.write(reinterpret_cast<char *>(&functionSpaceType),sizeof(int));      const DataTagged::DataMapType& lookup_1=tmp_1->getTagLookup();
2874    archiveFile.write(reinterpret_cast<char *>(&dataPointRank),sizeof(int));      for (i=lookup_0.begin();i!=lookup_0.end();i++) {
2875    archiveFile.write(reinterpret_cast<char *>(&dataPointSize),sizeof(int));        tmp_2->addTag(i->first); // use tmp_2 to get correct shape
2876    archiveFile.write(reinterpret_cast<char *>(&dataLength),sizeof(int));      }
2877    for (int dim = 0; dim < 4; dim++) {      for (i=lookup_1.begin();i!=lookup_1.end();i++) {
2878      archiveFile.write(reinterpret_cast<char *>(&flatShape[dim]),sizeof(int));        tmp_2->addTag(i->first);
2879    }      }
2880    for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {      // Compute an MVP for each tag
2881      archiveFile.write(reinterpret_cast<char *>(&referenceNumbers[sampleNo]),sizeof(int));      const DataTagged::DataMapType& lookup_2=tmp_2->getTagLookup();
2882    }      for (i=lookup_2.begin();i!=lookup_2.end();i++) {
2883    if (isTagged()) {        const double *ptr_0 = &(tmp_0->getDataByTagRO(i->first,0));
2884      for (int sampleNo=0; sampleNo<noSamples; sampleNo++) {        const double *ptr_1 = &(tmp_1->getDataByTagRO(i->first,0));
2885        archiveFile.write(reinterpret_cast<char *>(&tagNumbers[sampleNo]),sizeof(int));        double *ptr_2 = &(tmp_2->getDataByTagRW(i->first,0));
2886    
2887          matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
2888      }      }
   }  
2889    
   if (!archiveFile.good()) {  
     throw DataException("archiveData Error: problem writing to archive file");  
2890    }    }
2891      else if (arg_0_Z.isTagged()     && arg_1_Z.isExpanded()) {
2892    
2893    //      // After finding a common function space above the two inputs have the same numSamples and num DPPS
2894    // Archive underlying data values for each Data type      res = Data(0.0, shape2, arg_1_Z.getFunctionSpace(),true); // DataExpanded output
2895    int noValues;      DataTagged*   tmp_0=dynamic_cast<DataTagged*>(arg_0_Z.borrowData());
2896    switch (dataType) {      DataExpanded* tmp_1=dynamic_cast<DataExpanded*>(arg_1_Z.borrowData());
2897      case 0:      DataExpanded* tmp_2=dynamic_cast<DataExpanded*>(res.borrowData());
2898        // DataEmpty      if (tmp_0==0) { throw DataException("GTP Programming error - casting to DataTagged."); }
2899        noValues = 0;      if (tmp_1==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2900        archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));      if (tmp_2==0) { throw DataException("GTP Programming error - casting to DataExpanded."); }
2901        cout << "\tnoValues: " << noValues << endl;      int sampleNo_0,dataPointNo_0;
2902        break;      int numSamples_0 = arg_0_Z.getNumSamples();
2903      case 1:      int numDataPointsPerSample_0 = arg_0_Z.getNumDataPointsPerSample();
2904        // DataConstant      #pragma omp parallel for private(sampleNo_0,dataPointNo_0) schedule(static)
2905        noValues = m_data->getLength();      for (sampleNo_0 = 0; sampleNo_0 < numSamples_0; sampleNo_0++) {
2906        archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));        int offset_0 = tmp_0->getPointOffset(sampleNo_0,0); // They're all the same, so just use #0
2907        cout << "\tnoValues: " << noValues << endl;        const double *ptr_0 = &(arg_0_Z.getDataAtOffsetRO(offset_0));
2908        if (m_data->archiveData(archiveFile,noValues)) {        for (dataPointNo_0 = 0; dataPointNo_0 < numDataPointsPerSample_0; dataPointNo_0++) {
2909          throw DataException("archiveData Error: problem writing data to archive file");          int offset_1 = tmp_1->getPointOffset(sampleNo_0,dataPointNo_0);
2910        }          int offset_2 = tmp_2->getPointOffset(sampleNo_0,dataPointNo_0);
2911        break;          const double *ptr_1 = &(arg_1_Z.getDataAtOffsetRO(offset_1));
2912      case 2:          double *ptr_2 = &(res.getDataAtOffsetRW(offset_2));
2913        // DataTagged          matrix_matrix_product(SL, SM, SR, ptr_0, ptr_1, ptr_2, transpose);
       noValues = m_data->getLength();  
       archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));  
       cout << "\tnoValues: " << noValues << endl;  
       if (m_data->archiveData(archiveFile,noValues)) {  
         throw DataException("archiveData Error: problem writing data to archive file");  
2914        }        }
2915        break;      }
2916      case 3:  
2917        // DataExpanded    }
2918        noValues = m_data->getLength();    else if (arg_0_Z.isExpanded()   && arg_1_Z.isConstant()) {
2919        archiveFile.write(reinterpret_cast<char *>(&noValues),sizeof(int));