LCOV - code coverage report
Current view: top level - msvis/MSVis - AveragingTvi2.cc (source / functions) Hit Total Coverage
Test: casacpp_coverage.info Lines: 0 943 0.0 %
Date: 2024-10-10 11:40:37 Functions: 0 140 0.0 %

          Line data    Source code
       1             : #include <casacore/casa/Arrays/ArrayMath.h>
       2             : #include <casacore/casa/Arrays/ArrayPartMath.h>
       3             : #include <casacore/casa/BasicMath/Functors.h>
       4             : #include <casacore/casa/Utilities/CountedPtr.h>
       5             : #include <stdcasa/UtilJ.h>
       6             : #include <msvis/MSVis/AveragingTvi2.h>
       7             : #include <msvis/MSVis/AveragingVi2Factory.h>
       8             : #include <msvis/MSVis/MsRows.h>
       9             : #include <msvis/MSVis/VisBufferComponents2.h>
      10             : #include <msvis/MSVis/VisBuffer2.h>
      11             : #include <msvis/MSVis/VisBufferImpl2.h>
      12             : #include <msvis/MSVis/Vbi2MsRow.h>
      13             : #include <msvis/MSVis/VisibilityIterator2.h>
      14             : #include <tuple>
      15             : #include <set>
      16             : 
      17             : using std::set;
      18             : 
      19             : using namespace casacore;
      20             : namespace casa {
      21             : 
      22             : namespace vi {
      23             : 
      24             : namespace avg {
      25             : 
      26             : using casa::ms::MsRow;
      27             : 
      28             : 
      29             : ///////////////////////////////////////////////////////////
      30             : //
      31             : // VbAvg: Averaging VisBuffer
      32             : //
      33             : /*
      34             : 
      35             : Definition: A baseline sample (baseline for short) is a table row
      36             : with a particular (antenna1, antenna2) pair at a particular time.
      37             : 
      38             : 
      39             : Averaging does not cross chunk boundaries of the input VI so the
      40             : input VI determines what values are averaged together.  For example,
      41             : if the input VI is allows data from multiple scans into the same chunk
      42             : then some averaging across scans can occur; in this case the scan number
      43             : of the result will be the scan number of the first baseline instance
      44             : seen.
      45             : 
      46             : Row-level value treatment
      47             : =========================
      48             : 
      49             : The average is approximately computed on a per baseline basis:
      50             : 
      51             : averaged_baseline (antenna1, antenna2) =
      52             :     sumOverAveragingInterval (baseline (antenna1, antenna2)) /
      53             :     nBaselinesFoundInInterval
      54             : 
      55             : How row-level values are computed
      56             : ---------------------------------
      57             : 
      58             : Time - Set to time of first baseline making up the average plus
      59             :        half of the averaging interval.
      60             : Antenna1 - Copied from instance of baseline
      61             : Antenna2 - Copied from instance of baseline
      62             : Feed1 - Copied from instance of baseline
      63             : Feed2 - Copied from instance of baseline
      64             : Flag_Row - The flag row is the logical "and" of the flag rows
      65             :            averaged together.
      66             : Data_Desc_Id - Copied from instance of baseline
      67             : Processor_Id - Copied from instance of baseline
      68             : Field_Id - Copied from instance of baseline
      69             : Interval - Set to averaging interval
      70             : Exposure - Minimum of the interval and the sum of the exposures
      71             :            from unflagged rows
      72             : Time_Centroid - sum (timeCentroid[i] * exposure[i]) / sum (exposure[i])
      73             : Scan_Number - Copied from instance of baseline
      74             : Sigma - ???
      75             : Array_Id - Copied from instance of baseline
      76             : Observation_Id - Copied from instance of baseline
      77             : State_Id - Copied from instance of baseline
      78             : Uvw - Weighted average of the UVW values for the baseline
      79             : Weight - ???
      80             : 
      81             : Cube-level value treatment
      82             : --------------------------
      83             : 
      84             : For each baseline (i.e., antenna1, antenna2 pair) the average is
      85             : computed for each correlation (co) and channel (ch) of the data cube.
      86             : 
      87             : Data - sum (f(weightSpectrum (co, ch)) * data (co, ch)) /
      88             :        sum (f(weightSpectrum (co, ch)))
      89             :        f :== optional function applied to the weights; default is unity function.
      90             : Corrected_Data - Same was for Data however, VI setup can disable producing
      91             :                  averaged Corrected_Data
      92             : Model_Data - Same was for Data however, VI setup can disable producing
      93             :              averaged Model_Data
      94             : Weight_Spectrum - sum (weightSpectrum (co, ch))
      95             : Flag - Each averaged flag (correlation, channel) is the logical "and"
      96             :        of the flag (correlation, channel) making up the average.
      97             : 
      98             : */
      99             : 
     100             : 
     101             : class BaselineIndex {
     102             : 
     103             : public:
     104             : 
     105             :     // make noncopyable...
     106             :     BaselineIndex( const BaselineIndex& ) = delete;
     107             :     BaselineIndex& operator=( const BaselineIndex& ) = delete;
     108             : 
     109             :     BaselineIndex ();
     110             :     ~BaselineIndex ();
     111             : 
     112             :     Int operator () (Int antenna1, Int antenna2, Int spectralWindow);
     113             :     void configure (Int nAntennas, Int nSpw, const VisBuffer2 * vb);
     114             : 
     115             : 
     116             : private:
     117             : 
     118             :     enum {Empty = -1};
     119             : 
     120             :     class SpwIndex : public Matrix<Int>{
     121             : 
     122             :     public:
     123             : 
     124           0 :         SpwIndex (Int n) : Matrix<Int> (n, n, Empty) {}
     125             : 
     126             :         Int
     127           0 :         getBaselineNumber (Int antenna1, Int antenna2, Int & nBaselines)
     128             :         {
     129           0 :             Int i = (* this) (antenna1, antenna2);
     130             : 
     131           0 :             if (i == Empty){
     132             : 
     133           0 :                 i = nBaselines ++;
     134           0 :                 (* this) (antenna1, antenna2) = i;
     135             :             }
     136             : 
     137           0 :             return i;
     138             :         }
     139             : 
     140             :     private:
     141             : 
     142             :     };
     143             : 
     144             :     typedef vector<SpwIndex *> Index;
     145             : 
     146             :     SpwIndex * addSpwIndex (Int spw);
     147             :     Matrix<Int> * createMatrix ();
     148             :     void destroy();
     149             :     SpwIndex * getSpwIndex (Int spw);
     150             : 
     151             :     Index index_p;
     152             :     Int nAntennas_p;
     153             :     Int nBaselines_p;
     154             :     Int nSpw_p;
     155             : 
     156             : };
     157             : 
     158           0 : BaselineIndex::BaselineIndex ()
     159           0 : : nAntennas_p (0),
     160           0 :   nBaselines_p (0),
     161           0 :   nSpw_p (0)
     162           0 : {}
     163             : 
     164           0 : BaselineIndex::~BaselineIndex ()
     165             : {
     166           0 :     destroy();
     167           0 : }
     168             : 
     169             : Int
     170           0 : BaselineIndex::operator () (Int antenna1, Int antenna2, Int spectralWindow)
     171             : {
     172           0 :     SpwIndex * spwIndex = getSpwIndex (spectralWindow);
     173           0 :     if (spwIndex == 0){
     174           0 :         addSpwIndex (spectralWindow);
     175             :     }
     176             : 
     177           0 :     Int i = spwIndex->getBaselineNumber (antenna1, antenna2, nBaselines_p);
     178             : 
     179           0 :     return i;
     180             : }
     181             : 
     182             : 
     183             : 
     184             : BaselineIndex::SpwIndex *
     185           0 : BaselineIndex::addSpwIndex (Int i)
     186             : {
     187             :     // Delete an existing SpwIndex so that we start fresh
     188             : 
     189           0 :     delete index_p [i];
     190             : 
     191             :     // Create a new SpwIndex and insert it into the main index.
     192             : 
     193           0 :     index_p [i] = new SpwIndex (nAntennas_p);
     194             : 
     195           0 :     return index_p [i];
     196             : }
     197             : 
     198             : void
     199           0 : BaselineIndex::configure (Int nAntennas, Int nSpw, const VisBuffer2 * vb)
     200             : {
     201             :     // Capture the shape parameters
     202             : 
     203           0 :     nAntennas_p = nAntennas;
     204           0 :     nSpw_p = nSpw;
     205           0 :     nBaselines_p = 0;
     206             : 
     207             :     // Get rid of the existing index
     208             : 
     209           0 :     destroy ();
     210           0 :     index_p = Index (nSpw_p, (SpwIndex *) 0);
     211             : 
     212             :     // Fill out the index based on the contents of the first VB.
     213             :     // Usually this will determine the pattern for all of the VBs to be
     214             :     // averaged together so that is the ordering the index should
     215             :     // capture.
     216             : 
     217           0 :     for (rownr_t i = 0; i < vb->nRows(); i++){
     218             : 
     219             :         // Eagerly flesh out the Spw's index
     220             : 
     221           0 :         Int spw = vb->spectralWindows() (i);
     222           0 :         Int antenna1 = vb->antenna1()(i);
     223           0 :         Int antenna2 = vb->antenna2()(i);
     224             : 
     225           0 :         (* this) (antenna1, antenna2, spw);
     226             :     }
     227           0 : }
     228             : 
     229             : void
     230           0 : BaselineIndex::destroy ()
     231             : {
     232             :     // Delete all the dynamically allocated spectral window indices.
     233             :     // The vector destructor will take care of index_p itself.
     234             : 
     235           0 :     for (Index::iterator i = index_p.begin();
     236           0 :          i != index_p.end();
     237           0 :          i++){
     238             : 
     239           0 :         delete (* i);
     240             :     }
     241           0 : }
     242             : 
     243             : BaselineIndex::SpwIndex *
     244           0 : BaselineIndex::getSpwIndex (Int spw)
     245             : {
     246           0 :     AssertCc (spw < (int) index_p.size());
     247             : 
     248           0 :     SpwIndex * spwIndex = index_p [spw];
     249             : 
     250           0 :     if (spwIndex == 0){
     251           0 :         spwIndex = addSpwIndex (spw);
     252             :     }
     253             : 
     254           0 :     return spwIndex;
     255             : }
     256             : 
     257             : template <typename T>
     258             : class PrefilledMatrix {
     259             : 
     260             : public:
     261             : 
     262           0 :     PrefilledMatrix () : matrix_p (0, 0, 0), nChannels_p (0), nCorrelations_p (0) {}
     263             : 
     264             :     const Matrix<T> &
     265           0 :     getMatrix (Int nCorrelations, Int nChannels, const T & value)
     266             :     {
     267           0 :         if (nCorrelations != nCorrelations_p || nChannels != nChannels_p ||
     268           0 :             value != value_p){
     269             : 
     270           0 :             nCorrelations_p = nCorrelations;
     271           0 :             nChannels_p = nChannels;
     272           0 :             value_p = value;
     273             : 
     274           0 :             matrix_p.assign (Matrix<T> (nCorrelations_p, nChannels_p, value_p));
     275             :         }
     276             : 
     277           0 :         return matrix_p;
     278             :     }
     279             : 
     280             : private:
     281             : 
     282             :     Matrix<T> matrix_p;
     283             :     Int nChannels_p;
     284             :     Int nCorrelations_p;
     285             :     T value_p;
     286             : 
     287             : };
     288             : 
     289             : template <typename T>
     290             : class CachedPlaneAvg : public ms::CachedArrayBase {
     291             : 
     292             : public:
     293             : 
     294             : typedef const Cube<T> & (casa::vi::avg::VbAvg::* Accessor) () const;
     295             : 
     296           0 : CachedPlaneAvg (Accessor accessor) : accessor_p (accessor) {}
     297             : 
     298             : Matrix<T> &
     299           0 : getCachedPlane (casa::vi::avg::VbAvg * vb, Int row)
     300             : {
     301           0 :     if (! isCached()){
     302             : 
     303             :         //cache_p.reference ((vb ->* accessor_p)().xyPlane (row)); // replace with something more efficient
     304           0 :         referenceMatrix (cache_p, (vb ->* accessor_p)(), row);
     305           0 :         setCached ();
     306             :     }
     307             : 
     308           0 :     return cache_p;
     309             : }
     310             : 
     311             : private:
     312             : 
     313             :     static void
     314           0 :     referenceMatrix (Matrix<T> & cache, const Cube<T> & src, Int row)
     315             :     {
     316           0 :         IPosition shape = src.shape ();
     317           0 :         shape.resize (2);
     318             : 
     319             :         // This is a bit sleazy but it seems to be helpful to performance.
     320             :         // Assumes contiguously stored cube.
     321             : 
     322           0 :         T * storage = const_cast <T *> (& src (IPosition (3, 0, 0, row)));
     323             : 
     324           0 :         cache.takeStorage (shape, storage, casacore::SHARE);
     325           0 :     }
     326             : 
     327             :     Accessor accessor_p;
     328             :     Matrix<T> cache_p;
     329             : };
     330             : 
     331             : 
     332             : class VbAvg;
     333             : 
     334             : /**
     335             :  * Holds multiple rows in a buffer. changeRow() alternates between rows. The details
     336             :  * of how the rows are handled need to be traced to its parent class, Vbi2MsRow, and its
     337             :  * parent MsRow.
     338             :  */
     339             : class MsRowAvg : public ms::Vbi2MsRow {
     340             : 
     341             : public:
     342             : 
     343             :     MsRowAvg (rownr_t row, const VbAvg * vb);
     344             : 
     345             :     // Constructor for read/write access
     346             : 
     347             :     MsRowAvg (rownr_t row, VbAvg * vb);
     348             : 
     349           0 :     virtual ~MsRowAvg () {}
     350             : 
     351             :     Bool baselinePresent () const;
     352             :     /// For how many of the rows reachable by changeRow() does baselinePresent(() hold?
     353             :     /// That's equivalent to asking how many baselines are being
     354             :     Int nBaselinesPresent () const;
     355             : 
     356             :     Vector<Bool> correlationFlagsMutable ();
     357             :     const Matrix<Int> & counts () const;
     358             :     Int countsBaseline () const;
     359           0 :     Matrix<Bool> & flagsMutable () { return Vbi2MsRow::flagsMutable();}
     360             :     Double intervalLast () const;
     361             :     Double timeFirst () const;
     362             :     Double timeLast () const;
     363             :     Vector<Double> uvwFirst ();
     364             : 
     365             :     void setBaselinePresent (Bool isPresent);
     366             :     void setCounts (const Matrix<Int> &);
     367             :     void setCountsBaseline (Int);
     368             :     void setIntervalLast (Double);
     369             :     void setTimeFirst (Double);
     370             :     void setTimeLast (Double);
     371             : 
     372             :     Double getNormalizationFactor();
     373             :     void setNormalizationFactor(Double normalizationFactor);
     374             :     void accumulateNormalizationFactor(Double normalizationFactor);
     375             : 
     376             :     /**
     377             :      * For bookkeeping. Input row indices that have been added so far in the current row
     378             :      * (set with changeRow). This is a list of input rows attached to this averaged row,
     379             :      * needed to build the map of input->output row indices when this row is transferred to
     380             :      * the output/averaged buffer.
     381             :      */
     382             :     std::vector<Int> inputRowIdxs() { return inputRowIdxs_p[row()]; }
     383             :     /**
     384             :      * Adds into the 'inputRowIdxs' list the index of an input row from the buffer being
     385             :      * averaged.
     386             :      * @param idx the index of the input row in the input buffer
     387             :      */
     388           0 :     void addInputRowIdx(Int idx) {
     389           0 :         inputRowIdxs_p[row()].push_back(idx);
     390           0 :     }
     391             :     /**
     392             :      * Clear the list of input rows attached to this row. This is used once the row is
     393             :      * transferred to the output/averaged buffer (typically after every average interval).
     394             :      */
     395             :     void clearRowIdxs() { inputRowIdxs_p[row()].clear(); }
     396             : 
     397             : private:
     398             : 
     399             :     void configureCountsCache ();
     400             : 
     401             :     mutable CachedPlaneAvg<Int> countsCache_p;
     402             :     Vector<Double> normalizationFactor_p;
     403             :     VbAvg * vbAvg_p; // [use]
     404             :     // map: input buffer row index -> output buffer row index
     405             :     std::map<Int, std::vector<Int>> inputRowIdxs_p;
     406             : };
     407             : 
     408             : /**
     409             :  * It looks like the intended usage of this VbAvg class (from AveragingTvi2::
     410             :  * produceSubchunk()) is as follows:
     411             :  *
     412             :  * // Use a "VbAvg vbAvg":
     413             :  * VisBufferImpl2 * vbToFill = // get/create output (averaged) buffer
     414             :  * vbToFill->setFillable(true);
     415             :  * vbAvg.setBufferToFill(vbToFill);
     416             :  * // we have the input buffer (to be averaged) in "VisibilityIteratorImpl2 vii;
     417             :  * while (vii->more()) {
     418             :  *    ...
     419             :  *    vbAvg.accumulate(vb, subhunk);
     420             :  * }
     421             :  * vbAvg.finalizeAverages();
     422             :  * vbAvg.finalizeBufferFillnig();
     423             :  */
     424             : class VbAvg : public VisBufferImpl2 {
     425             : 
     426             : public:
     427             : 
     428             :     friend class MsRowAvg;
     429             : 
     430             :     VbAvg (const AveragingParameters & averagingParameters, ViImplementation2 * vi);
     431             : 
     432             :     void accumulate (const VisBuffer2 * vb, const Subchunk & subchunk);
     433             :     const Cube<Int> & counts () const;
     434             :     Bool empty () const;
     435             :     void finalizeBufferFilling ();
     436             :     void finalizeAverages ();
     437             :     MsRowAvg * getRow (Int row) const;
     438             :     MsRowAvg * getRowMutable (Int row);
     439             :     Bool isComplete () const;
     440             :     Bool isUsingUvwDistance () const;
     441             :     void markEmpty ();
     442             :     int nSpectralWindowsInBuffer () const;
     443             :     void setBufferToFill (VisBufferImpl2 *);
     444             :     void startChunk (ViImplementation2 *);
     445             :     Int getBaselineIndex (Int antenna1, Int antenna2, Int spw) const;
     446             :     // Vector with row idx in the averaged/ooutput buffers
     447           0 :     const std::vector<size_t> & row2AvgRow() const { return row2AvgRow_p; };
     448             : 
     449             : protected:
     450             : 
     451             :     class Doing {
     452             :     public:
     453             : 
     454           0 :         Doing ()
     455           0 :         : correctedData_p (false),
     456           0 :           modelData_p (false),
     457           0 :           observedData_p (false),
     458           0 :           floatData_p(false),
     459           0 :           onlymetadata_p(true),
     460           0 :           weightSpectrumIn_p (false),
     461           0 :           sigmaSpectrumIn_p (false),
     462           0 :           weightSpectrumOut_p (false),
     463           0 :           sigmaSpectrumOut_p (false)
     464           0 :         {}
     465             : 
     466             :         Bool correctedData_p;
     467             :         Bool modelData_p;
     468             :         Bool observedData_p;
     469             :         Bool floatData_p;
     470             :         Bool onlymetadata_p;
     471             :         Bool weightSpectrumIn_p;
     472             :         Bool sigmaSpectrumIn_p;
     473             :         Bool weightSpectrumOut_p;
     474             :         Bool sigmaSpectrumOut_p;
     475             :     };
     476             : 
     477             :     class AccumulationParameters {
     478             : 
     479             :     public:
     480             : 
     481           0 :         AccumulationParameters (MsRow * rowInput, MsRowAvg * rowAveraged, const Doing & doing)
     482           0 :         : correctedIn_p (doing.correctedData_p ? rowInput->corrected().data() : 0),
     483           0 :           correctedOut_p (doing.correctedData_p ? rowAveraged->correctedMutable ().data(): 0),
     484           0 :           flagCubeIn_p (rowInput->flags().data()),
     485           0 :           flagCubeOut_p (rowAveraged->flagsMutable().data()),
     486           0 :           floatDataIn_p (doing.floatData_p ? rowInput->singleDishData().data() : 0),
     487           0 :           floatDataOut_p (doing.floatData_p ? rowAveraged->singleDishDataMutable().data() : 0),
     488           0 :           modelIn_p (doing.modelData_p ? rowInput->model().data(): 0),
     489           0 :           modelOut_p (doing.modelData_p ? rowAveraged->modelMutable().data() : 0),
     490           0 :           observedIn_p (doing.observedData_p ? rowInput->observed().data() : 0),
     491           0 :           observedOut_p (doing.observedData_p ? rowAveraged->observedMutable().data() : 0),
     492           0 :           sigmaIn_p (& rowInput->sigma()),
     493           0 :           sigmaOut_p (& rowAveraged->sigmaMutable()),
     494           0 :           sigmaSpectrumIn_p (doing.sigmaSpectrumIn_p ? rowInput->sigmaSpectrum().data() : 0),
     495           0 :           sigmaSpectrumOut_p (doing.sigmaSpectrumOut_p ? rowAveraged->sigmaSpectrumMutable().data() : 0),
     496           0 :           weightIn_p (& rowInput->weight()),
     497           0 :           weightOut_p (& rowAveraged->weightMutable()),
     498           0 :           weightSpectrumIn_p (doing.weightSpectrumIn_p ? rowInput->weightSpectrum().data() : 0),
     499           0 :           weightSpectrumOut_p (doing.weightSpectrumOut_p ? rowAveraged->weightSpectrumMutable().data() : 0)
     500           0 :         {}
     501             : 
     502           0 :         void incrementCubePointers()
     503             :         {
     504             :             // For improved performance this class is designed to sweep the cube data elements in this row
     505             :             // in the order (correlation0, channel0), (correlation1, channel1), etc.
     506             : 
     507           0 :             correctedIn_p && correctedIn_p ++;
     508           0 :             correctedOut_p && correctedOut_p ++;
     509           0 :             flagCubeIn_p && flagCubeIn_p ++;
     510           0 :             flagCubeOut_p && flagCubeOut_p ++;
     511           0 :             floatDataIn_p && floatDataIn_p ++;
     512           0 :             floatDataOut_p && floatDataOut_p ++;
     513           0 :             modelIn_p && modelIn_p ++;
     514           0 :             modelOut_p && modelOut_p ++;
     515           0 :             observedIn_p && observedIn_p ++;
     516           0 :             observedOut_p && observedOut_p ++;
     517           0 :             sigmaSpectrumIn_p && sigmaSpectrumIn_p ++;
     518           0 :             sigmaSpectrumOut_p && sigmaSpectrumOut_p ++;
     519           0 :             weightSpectrumIn_p && weightSpectrumIn_p ++;
     520           0 :             weightSpectrumOut_p && weightSpectrumOut_p ++;
     521           0 :         }
     522             : 
     523             :         inline const Complex *
     524           0 :         correctedIn ()
     525             :         {
     526           0 :             assert (correctedIn_p != 0);
     527           0 :             return correctedIn_p;
     528             :         }
     529             : 
     530             :         inline Complex *
     531           0 :         correctedOut ()
     532             :         {
     533           0 :             assert (correctedOut_p != 0);
     534           0 :             return correctedOut_p;
     535             :         }
     536             : 
     537             :         inline const Float *
     538           0 :         floatDataIn ()
     539             :         {
     540           0 :             return floatDataIn_p;
     541             :         }
     542             : 
     543             :         inline Float *
     544           0 :         floatDataOut ()
     545             :         {
     546           0 :             return floatDataOut_p;
     547             :         }
     548             : 
     549             :         inline const Complex *
     550           0 :         modelIn ()
     551             :         {
     552           0 :             assert (modelIn_p != 0);
     553           0 :             return modelIn_p;
     554             :         }
     555             : 
     556             :         inline Complex *
     557           0 :         modelOut ()
     558             :         {
     559           0 :             assert (modelOut_p != 0);
     560           0 :             return modelOut_p;
     561             :         }
     562             : 
     563             :         inline const Complex *
     564           0 :         observedIn ()
     565             :         {
     566           0 :             assert (observedIn_p != 0);
     567           0 :             return observedIn_p;
     568             :         }
     569             : 
     570             :         inline Complex *
     571           0 :         observedOut ()
     572             :         {
     573           0 :             assert (observedOut_p != 0);
     574           0 :             return observedOut_p;
     575             :         }
     576             : 
     577             :         inline const Bool *
     578             :         flagCubeIn ()
     579             :         {
     580             :             assert (flagCubeIn_p != 0);
     581             :             return flagCubeIn_p;
     582             :         }
     583             : 
     584             :         inline Bool *
     585             :         flagCubeOut ()
     586             :         {
     587             :             assert (flagCubeOut_p != 0);
     588             :             return flagCubeOut_p;
     589             :         }
     590             : 
     591             :         inline const Float *
     592           0 :         sigmaSpectrumIn ()
     593             :         {
     594           0 :             return sigmaSpectrumIn_p;
     595             :         }
     596             : 
     597             :         inline Float *
     598           0 :         sigmaSpectrumOut ()
     599             :         {
     600           0 :             assert (sigmaSpectrumOut_p != 0);
     601           0 :             return sigmaSpectrumOut_p;
     602             :         }
     603             : 
     604             :         inline const Float *
     605           0 :         weightSpectrumIn ()
     606             :         {
     607           0 :             assert (weightSpectrumIn_p != 0);
     608           0 :             return weightSpectrumIn_p;
     609             :         }
     610             : 
     611             :         inline Float *
     612           0 :         weightSpectrumOut ()
     613             :         {
     614           0 :             assert (weightSpectrumOut_p != 0);
     615           0 :             return weightSpectrumOut_p;
     616             :         }
     617             : 
     618             :         inline const Vector<Float> &
     619             :         weightIn ()
     620             :         {
     621             :             assert (weightIn_p != 0);
     622             :             return *weightIn_p;
     623             :         }
     624             : 
     625             :         inline Vector<Float> &
     626             :         weightOut ()
     627             :         {
     628             :             assert (weightOut_p != 0);
     629             :             return *weightOut_p;
     630             :         }
     631             : 
     632             :         inline const Vector<Float> &
     633             :         sigmaIn ()
     634             :         {
     635             :             assert (sigmaIn_p != 0);
     636             :             return *sigmaIn_p;
     637             :         }
     638             : 
     639             :         inline Vector<Float> &
     640             :         sigmaOut ()
     641             :         {
     642             :             assert (sigmaOut_p != 0);
     643             :             return *sigmaOut_p;
     644             :         }
     645             : 
     646             : 
     647             : 
     648             :     private:
     649             : 
     650             :         const Complex * correctedIn_p;
     651             :         Complex * correctedOut_p;
     652             :         const Bool * flagCubeIn_p;
     653             :         Bool * flagCubeOut_p;
     654             :         const Float * floatDataIn_p;
     655             :         Float * floatDataOut_p;
     656             :         const Complex * modelIn_p;
     657             :         Complex * modelOut_p;
     658             :         const Complex * observedIn_p;
     659             :         Complex * observedOut_p;
     660             :         const Vector<Float> * sigmaIn_p;
     661             :         Vector<Float> * sigmaOut_p;
     662             :         const Float * sigmaSpectrumIn_p;
     663             :         Float * sigmaSpectrumOut_p;
     664             :         const Vector<Float> * weightIn_p;
     665             :         Vector<Float> * weightOut_p;
     666             :         const Float * weightSpectrumIn_p;
     667             :         Float * weightSpectrumOut_p;
     668             : 
     669             : 
     670             :     };
     671             : 
     672             :     std::pair<Bool, Vector<Double> > accumulateCubeData (MsRow * rowInput, MsRowAvg * rowAveraged);
     673             :     void accumulateElementForCubes (AccumulationParameters & accumulationParameters,
     674             :                                     Bool zeroAccumulation);
     675             :     template<typename T>
     676             :     void
     677             :     accumulateElementForCube (const T * unweightedValue,
     678             :                               Float weight,
     679             :                               Bool zeroAccumulation,
     680             :                               T * accumulator);
     681             : 
     682             :     template <typename T>
     683             :     T accumulateRowDatum (const T & averagedValue, const T & inputValue,
     684             :                           Bool resetAverage);
     685             : 
     686             : 
     687             : 
     688             :     void accumulateExposure (const VisBuffer2 *);
     689             :     /*
     690             :      * Called once per row in the input buffer
     691             :      * @param rowInput row from the input buffer being averaged
     692             :      * @param rowAveraged changing "accumulator" row for the output buffer
     693             :      * @param subchunk - hard to understand
     694             :      * @param iidx index of the input row in the input buffer
     695             :      */
     696             :     void accumulateOneRow (MsRow * rowInput, MsRowAvg * rowAveraged,
     697             :                            const Subchunk & subchunk, Int iidx);
     698             :     void accumulateRowData (MsRow * rowInput, MsRowAvg * rowAveraged, Double adjustedWeight,
     699             :                             Bool rowFlagged);
     700             :     void accumulateTimeCentroid (const VisBuffer2 * input);
     701             :     void captureIterationInfo (VisBufferImpl2 * dstVb, const VisBuffer2 * srcVb,
     702             :                                const Subchunk & subchunk);
     703             :     void copyBaseline (Int sourceBaseline, Int targetBaseline);
     704             :     template<typename T>
     705             :     void copyBaselineScalar (Int sourceBaseline, Int targetBaseline,
     706             :                              Vector<T> & columnVector);
     707             :     template<typename T>
     708             :     void copyCubePlaneIf (Bool condition, Int sourceBaseline,
     709             :                           Int targetBaseline, Cube<T> & cube);
     710             :     void copyIdValues (MsRow * rowInput, MsRowAvg * rowAveraged);
     711             :     void copyIdValue (Int inputId, Int & averagedId);
     712             :     void finalizeBaseline (MsRowAvg *);
     713             :     void finalizeBaselineIfNeeded (MsRow * rowInput, MsRowAvg * rowAveraged,
     714             :                                    const Subchunk & subchunk);
     715             :     void finalizeCubeData (MsRowAvg *);
     716             :     void finalizeRowData (MsRowAvg *);
     717             :     AccumulationParameters * getAccumulationParameters (MsRow * rowInput,
     718             :                                                         MsRowAvg * rowAveraged);
     719             :     Int getBaselineIndex (const MsRow *) const;
     720             :     void initializeBaseline (MsRow * rowInput, MsRowAvg * rowAveraged,
     721             :                              const Subchunk & subchunk);
     722             :     Int nBaselines () const;
     723             :     void prepareIds (const VisBuffer2 * vb);
     724             :     void removeMissingBaselines ();
     725             : 
     726             : private:
     727             : 
     728             :     void setupVbAvg (const VisBuffer2 *);
     729             :     void setupArrays (Int nCorrelations, Int nChannels, Int nBaselines);
     730             :     void setupBaselineIndices (Int nAntennas, const VisBuffer2 * vb);
     731             : 
     732             :     /// A baseline being averaged into a MSRowAvg is put into the output/averaged buffer and
     733             :     /// set as not present
     734             :     void transferBaseline (MsRowAvg *);
     735             : 
     736             :     template <typename T>
     737             :     static T
     738           0 :     distanceSquared (const Vector<T> & p1, const Vector<T> & p2)
     739             :     {
     740           0 :         assert (p1.size() == 3 && p2.size() == 3);
     741             : 
     742           0 :         T distanceSquared = 0;
     743             : 
     744           0 :         for (Int i = 0; i < 3; i++){
     745           0 :             T delta = p1[i] - p2[i];
     746           0 :             distanceSquared += delta * delta;
     747             :         }
     748             : 
     749           0 :         return distanceSquared;
     750             :     }
     751             : 
     752             :     Double averagingInterval_p;
     753             :     AveragingOptions averagingOptions_p;
     754             :     const ViImplementation2 * averagingVii_p;
     755             :     mutable BaselineIndex baselineIndex_p; // map of antenna1,antenna2 to row number in this VB.
     756             :     Vector<Bool> baselinePresent_p; // indicates whether baseline has any data
     757             :     Vector<Double> normalizationFactor_p; // indicates whether baseline has any data
     758             :     VisBufferImpl2 * bufferToFill_p;
     759             :     Bool complete_p; // average is complete
     760             :     Matrix<Bool> correlationFlags_p; // used for weight accumulation
     761             :     Cube<Int> counts_p; // number of items summed together for each correlation, channel, baseline item.
     762             :     Vector<Int> countsBaseline_p; // number of items summed together for each baseline.
     763             :     Doing doing_p;
     764             :     Bool empty_p; // true when buffer hasn't seen any data
     765             :     Vector<Double> intervalLast_p;
     766             :     Double maxTimeDistance_p;
     767             :     Double maxUvwDistanceSquared_p;
     768             :     Bool needIterationInfo_p;
     769             :     VisBufferComponents2 optionalComponentsToCopy_p;
     770             :     Int rowIdGenerator_p;
     771             :     Double sampleInterval_p;
     772             :     Double startTime_p; // time of the first sample in average
     773             :     Vector<Double> timeFirst_p;
     774             :     Vector<Double> timeLast_p;
     775             :     mutable PrefilledMatrix<Bool> trueBool_p;
     776             :     Matrix<Double> uvwFirst_p;
     777             :     Bool usingUvwDistance_p;
     778             :     mutable PrefilledMatrix<Int> zeroInt_p;
     779             :     mutable PrefilledMatrix<Float> zeroFloat_p;
     780             : 
     781             :     std::vector<size_t> row2AvgRow_p;
     782             : 
     783             :     LogIO logger_p;
     784             : };
     785             : 
     786             : ///////////////////////////////////////////////////////////
     787             : //
     788             : // Set of Averaging VisBuffers, one per active DD ID
     789             : 
     790             : 
     791             : //class VbSet {
     792             : //
     793             : //public:
     794             : //
     795             : //    VbSet (const AveragingParameters & averagingParameters);
     796             : //    ~VbSet ();
     797             : //
     798             : //    void accumulate (const VisBuffer2 *, const Subchunk & subchunk);
     799             : //    void finalizeBufferFilling ();
     800             : //    void setBufferToFill (VisBuffer2 *);
     801             : //
     802             : //
     803             : //    VbAvg * add (Int ddId);
     804             : //    Bool anyAveragesReady(Int ddid = -1) const;
     805             : //    Bool contains (Int ddId) const;
     806             : ////    void finalizeAverage (Int ddId);
     807             : //    void finalizeAverages ();
     808             : //    void finalizeRowIfNeeded (ms::MsRow * rowInput, avg::MsRowAvg * rowAveraged, const Subchunk & subchunk);
     809             : //    void flush (Bool okIfNonempty = false, ViImplementation2 * vi = 0); // delete all averagers
     810             : //    Int getFirstReadyDdid () const;
     811             : //    void transferAverage (Int ddId, VisBuffer2 * vb);
     812             : //    Bool vbPastAveragingInterval (const VisBuffer2 * vb) const;
     813             : //    void zero ();
     814             : //
     815             : //protected:
     816             : //
     817             : //    void seeIfCubeColumnsExist (ViImplementation2 * vi);
     818             : //
     819             : //private:
     820             : //
     821             : //    typedef map<Int, VbAvg *> Averagers;
     822             : //
     823             : //    const Double averagingInterval_p;
     824             : //    AveragingOptions averagingOptions_p;
     825             : //    const AveragingParameters averagingParameters_p;
     826             : //    Bool doingCorrectedData_p;
     827             : //    Bool doingModelData_p;
     828             : //    Bool doingObservedData_p;
     829             : //    Bool doingWeightSpectrum_p;
     830             : //    Bool doingsigmaSpectrum_p;
     831             : //    Averagers vbAveragers_p;
     832             : //};
     833             : 
     834           0 : MsRowAvg::MsRowAvg (rownr_t row, const VbAvg * vb)
     835             : : Vbi2MsRow (row, vb),
     836           0 :   countsCache_p (& VbAvg::counts),
     837           0 :   normalizationFactor_p(0.0),
     838           0 :   vbAvg_p (const_cast<VbAvg *> (vb))
     839             : {
     840           0 :     configureCountsCache();
     841           0 : }
     842             : 
     843             : 
     844             : // Constructor for read/write access
     845             : 
     846           0 : MsRowAvg::MsRowAvg (rownr_t row, VbAvg * vb)
     847             : : Vbi2MsRow (row, vb),
     848           0 :   countsCache_p (& VbAvg::counts),
     849           0 :   normalizationFactor_p(0.0),
     850           0 :   vbAvg_p (vb)
     851             : {
     852           0 :     configureCountsCache();
     853           0 : }
     854             : 
     855             : Bool
     856           0 : MsRowAvg::baselinePresent () const
     857             : {
     858           0 :     return vbAvg_p->baselinePresent_p (row ());
     859             : }
     860             : 
     861             : Int
     862           0 : MsRowAvg::nBaselinesPresent () const
     863             : {
     864           0 :     return std::count(vbAvg_p->baselinePresent_p.begin(), vbAvg_p->baselinePresent_p.end(),
     865           0 :                       true);
     866             : }
     867             : 
     868             : void
     869           0 : MsRowAvg::configureCountsCache ()
     870             : {
     871           0 :     addToCachedArrays (countsCache_p);
     872           0 : }
     873             : 
     874             : const Matrix<Int> &
     875           0 : MsRowAvg::counts () const
     876             : {
     877           0 :     return countsCache_p.getCachedPlane (dynamic_cast<VbAvg *> (getVbi()), row());
     878             : }
     879             : 
     880             : Vector<Bool>
     881           0 : MsRowAvg::correlationFlagsMutable ()
     882             : {
     883           0 :     return vbAvg_p->correlationFlags_p.column (row());
     884             : }
     885             : 
     886             : Int
     887           0 : MsRowAvg::countsBaseline () const
     888             : {
     889           0 :     return vbAvg_p->countsBaseline_p (row ());
     890             : }
     891             : 
     892             : void
     893           0 : MsRowAvg::setBaselinePresent (Bool value)
     894             : {
     895           0 :     vbAvg_p->baselinePresent_p (row ()) = value;
     896           0 : }
     897             : 
     898             : 
     899             : void
     900           0 : MsRowAvg::setCountsBaseline (Int value)
     901             : {
     902           0 :     vbAvg_p->countsBaseline_p (row ()) = value;
     903           0 : }
     904             : 
     905             : Double
     906           0 : MsRowAvg::intervalLast () const
     907             : {
     908           0 :     return vbAvg_p->intervalLast_p (row ());
     909             : }
     910             : 
     911             : 
     912             : Double
     913           0 : MsRowAvg::timeFirst () const
     914             : {
     915           0 :     return vbAvg_p->timeFirst_p (row ());
     916             : }
     917             : 
     918             : Double
     919           0 : MsRowAvg::timeLast () const
     920             : {
     921           0 :     return vbAvg_p->timeLast_p (row ());
     922             : }
     923             : 
     924             : Vector<Double>
     925           0 : MsRowAvg::uvwFirst ()
     926             : {
     927           0 :     return vbAvg_p->uvwFirst_p.column (row());
     928             : }
     929             : 
     930             : 
     931             : void
     932           0 : MsRowAvg::setCounts (const Matrix<Int> & value)
     933             : {
     934           0 :     Matrix<Int> & theCounts = countsCache_p.getCachedPlane (dynamic_cast<VbAvg *> (getVbi()), row());
     935           0 :     theCounts = value;
     936           0 : }
     937             : 
     938             : void
     939           0 : MsRowAvg::setIntervalLast (Double value)
     940             : {
     941           0 :     vbAvg_p->intervalLast_p (row ()) = value;
     942           0 : }
     943             : 
     944             : 
     945             : void
     946           0 : MsRowAvg::setTimeFirst (Double value)
     947             : {
     948           0 :     vbAvg_p->timeFirst_p (row ()) = value;
     949           0 : }
     950             : 
     951             : void
     952           0 : MsRowAvg::setTimeLast (Double value)
     953             : {
     954           0 :     vbAvg_p->timeLast_p (row ()) = value;
     955           0 : }
     956             : 
     957           0 : Double MsRowAvg::getNormalizationFactor()
     958             : {
     959           0 :         return vbAvg_p->normalizationFactor_p (row ());
     960             : }
     961             : 
     962           0 : void MsRowAvg::setNormalizationFactor(Double normalizationFactor)
     963             : {
     964           0 :         vbAvg_p->normalizationFactor_p (row ()) = normalizationFactor;
     965           0 : }
     966             : 
     967           0 : void MsRowAvg::accumulateNormalizationFactor(Double normalizationFactor)
     968             : {
     969           0 :         vbAvg_p->normalizationFactor_p (row ()) += normalizationFactor;
     970           0 : }
     971             : 
     972           0 : VbAvg::VbAvg (const AveragingParameters & averagingParameters, ViImplementation2 * vii)
     973             : : VisBufferImpl2 (vii, VbRekeyable),
     974           0 :   averagingInterval_p (averagingParameters.getAveragingInterval ()),
     975           0 :   averagingOptions_p (averagingParameters.getOptions()),
     976           0 :   averagingVii_p (vii),
     977           0 :   bufferToFill_p (0),
     978           0 :   complete_p (false),
     979           0 :   doing_p (), // all false until determined later on
     980           0 :   empty_p (true),
     981           0 :   maxTimeDistance_p (averagingParameters.getAveragingInterval() * (0.999)),
     982             :         // Shrink it just a bit for roundoff
     983           0 :   maxUvwDistanceSquared_p (pow(averagingParameters.getMaxUvwDistance(),2)),
     984           0 :   needIterationInfo_p (true),
     985           0 :   rowIdGenerator_p (0),
     986           0 :   sampleInterval_p (0),
     987           0 :   startTime_p (0),
     988           0 :   usingUvwDistance_p (averagingParameters.getOptions().contains (AveragingOptions::BaselineDependentAveraging))
     989           0 : {}
     990             : 
     991             : /**
     992             :  * Calculates the row index in the output buffer, given an averaged row, a baseline index
     993             :  * corresponding to this averaged row, and the magic "rowIdGenerator" of the VbAvg.
     994             :  *
     995             :  * @param nrows number of rows in the input buffer being averaged
     996             :  * @param baselineIndex index of the baseline being averaged into the rowAveraged
     997             :  * @param rowAveraged "accumulator" row being produced for the output buffer
     998             :  * @param rowIdGen the rowIdGenerator of the VbAvg, which increases (in a not so clean way)
     999             :  *        for every new baseline inside every chunk
    1000             :  */
    1001             : Int
    1002           0 : calcOutRowIdx(Int nrows, Int baselineIndex, const MsRowAvg* rowAveraged, Int rowIdGen)
    1003             : {
    1004           0 :     auto nBasePresent = rowAveraged->nBaselinesPresent();
    1005             :     // check whether multiple time steps are being averaged into the output buffer
    1006             :     // (row blocking or similar feature is enabled)
    1007           0 :     const bool multitime =  nrows > nBasePresent;
    1008             : 
    1009           0 :     if (!multitime) {
    1010             :         // There is only one time step, so the index must be simply the baseline index.
    1011             :         // with row blocking disabled, it doesn't seem to be possible to make sense out of
    1012             :         // rowIdgenerator_p for the purpose of this calculation -> skip the more general
    1013             :         // calculation from below and just use baseline index.
    1014           0 :         return baselineIndex;
    1015             :     }
    1016             : 
    1017             :     // the rowIdgenerator_p that we get in rowIdGen increases +1 for every new baseline.
    1018             :     // It is not really a proper (input) row id. After all baselines have been seen for a
    1019             :     // time step, the rows of the next time steps will get the same id.
    1020             :     // And to turn it into a valid output id we need the following
    1021             :     // "divide_with_roundup + multiply + baseline_index"
    1022           0 :     Int rowIdG_div_baselines_roundup = 0;
    1023           0 :     if (nBasePresent > 0)
    1024           0 :         rowIdG_div_baselines_roundup = (rowIdGen + nBasePresent - 1)/ nBasePresent;
    1025           0 :     const Int id = rowIdG_div_baselines_roundup * nBasePresent + baselineIndex;
    1026             : 
    1027           0 :     return id;
    1028             : }
    1029             : 
    1030             : void
    1031           0 : VbAvg::accumulate (const VisBuffer2 * vb, const Subchunk & subchunk)
    1032             : {
    1033           0 :     if (empty_p){
    1034           0 :         setupVbAvg (vb);
    1035             :     }
    1036             : 
    1037           0 :     if (needIterationInfo_p){
    1038           0 :         captureIterationInfo (bufferToFill_p, vb, subchunk);
    1039           0 :         needIterationInfo_p = false;
    1040             :     }
    1041             : 
    1042           0 :     assert (bufferToFill_p != 0);
    1043             : 
    1044           0 :     MsRowAvg * rowAveraged = getRowMutable (0);
    1045           0 :     MsRow * rowInput = vb->getRow (0);
    1046             : 
    1047           0 :     auto nrows = vb->nRows();
    1048           0 :     row2AvgRow_p.resize(nrows);
    1049           0 :     for (rownr_t row = 0; row < nrows; ++row){
    1050             : 
    1051           0 :         rowInput->changeRow (row);
    1052           0 :         Int baselineIndex = getBaselineIndex (rowInput);
    1053             : 
    1054           0 :         rowAveraged->changeRow (baselineIndex);
    1055             : 
    1056           0 :         accumulateOneRow (rowInput, rowAveraged, subchunk, row);
    1057             : 
    1058           0 :         row2AvgRow_p[row] = calcOutRowIdx(nrows, baselineIndex, rowAveraged,
    1059             :                                           rowIdGenerator_p);
    1060             :     }
    1061             : 
    1062           0 :     delete rowAveraged;
    1063           0 :     delete rowInput;
    1064             : 
    1065           0 : }
    1066             : 
    1067             : void
    1068           0 : VbAvg::accumulateOneRow (MsRow * rowInput, MsRowAvg * rowAveraged, const Subchunk & subchunk,
    1069             :                          Int iidx)
    1070             : {
    1071           0 :     finalizeBaselineIfNeeded (rowInput, rowAveraged, subchunk);
    1072             : 
    1073           0 :     if (! rowAveraged->baselinePresent())
    1074             :     {
    1075             : 
    1076           0 :         initializeBaseline (rowInput, rowAveraged, subchunk);
    1077             :     }
    1078             : 
    1079             :     // bookkeeping - save for later that this input row is being averaged into the output row
    1080           0 :     rowAveraged->addInputRowIdx(iidx);
    1081             : 
    1082             :     // Accumulate data that is matrix-valued (e.g., vis, visModel, etc.).
    1083             :     // The computation of time centroid requires the use of the weight column
    1084             :     // adjusted for the flag cube.  Get the before and after weight accumulation
    1085             :     // and the difference is the adjusted weight for this row.
    1086             : 
    1087           0 :     Vector<Double> adjustedWeights;
    1088           0 :     Bool rowFlagged = false;
    1089             : 
    1090           0 :     std::tie (rowFlagged, adjustedWeights) = accumulateCubeData (rowInput, rowAveraged);
    1091             : 
    1092           0 :     Double adjustedWeight = 0;
    1093           0 :     for (Int c = 0; c < nCorrelations(); c++){
    1094             : 
    1095           0 :         adjustedWeight += adjustedWeights (c);
    1096             :     }
    1097             : 
    1098             :     // Accumulate the non matrix-valued data
    1099           0 :     accumulateRowData (rowInput, rowAveraged, adjustedWeight, rowFlagged);
    1100           0 : }
    1101             : 
    1102             : //void
    1103             : //VbAvg::accumulate (const VisBuffer2 * vb, const Subchunk & subchunk)
    1104             : //{
    1105             : //    // "Add" the contents of this buffer to the accumulation.
    1106             : //
    1107             : //    if (empty_p){
    1108             : //
    1109             : //        // Initialize the buffer if this is the first time bit of data that it is
    1110             : //        // being used after either creation or clearing.
    1111             : //
    1112             : //        prepareForFirstData (vb, subchunk);
    1113             : //
    1114             : //        empty_p = false;
    1115             : //    }
    1116             : //
    1117             : //    // Averaging can be computed as flagged or unflagged.  If all the inputs to a
    1118             : //    // datum are flagged, then the averaged datum (e.g., a visibility point)
    1119             : //    // will also be flagged.  For an unflagged averaged datum, it will represent
    1120             : //    // the average of all of the unflagged points for that datum.  This is done
    1121             : //    // by assuming the accumulation is flagged and continuing to accumulate
    1122             : //    // flagged points until the first unflagged point for a datum is encountered;
    1123             : //    // when this happens the count is zeroed and the averaged datum's flag is cleared.
    1124             : //
    1125             : //    // Loop over all of the rows in the VB.  Map each one to a baseline and then accumulate
    1126             : //    // the values for each correlation,channel cell.  Each row in the accumulating VB corresponds
    1127             : //    // to one baseline (i.e., pair of (antenna1, antenna2) where antenna1 <= antenna2).
    1128             : //
    1129             : //    ThrowIf (vb->nRows() > nBaselines(),
    1130             : //             String::format ("Expected %d baselines in VisBuffer but it contained %d rows",
    1131             : //                             nBaselines(), nRows()));
    1132             : //
    1133             : //    for (Int row = 0; row < vb->nRows(); row ++){
    1134             : //
    1135             : //        // Accumulate data for fields that are scalars (and uvw) in each row
    1136             : //
    1137             : //        accumulateRowData (vb, row);
    1138             : //
    1139             : //        // Accumulate data that is matrix-valued (e.g., vis, visModel, etc.)
    1140             : //
    1141             : //        accumulateCubeData (vb, row);
    1142             : //    }
    1143             : //}
    1144             : 
    1145             : void
    1146           0 : VbAvg::finalizeBufferFilling ()
    1147             : {
    1148           0 :     bufferToFill_p->appendRowsComplete();
    1149           0 :     bufferToFill_p = 0; // decouple
    1150           0 : }
    1151             : 
    1152             : template<typename T>
    1153             : inline void
    1154           0 : VbAvg::accumulateElementForCube (const T * unweightedValue,
    1155             :                                  Float weight,
    1156             :                                  Bool zeroAccumulation,
    1157             :                                  T * accumulator)
    1158             : {
    1159             :     // Update the sum for this model visibility cube element.
    1160             : 
    1161           0 :     if (zeroAccumulation){
    1162           0 :         * accumulator = (* unweightedValue) * weight;
    1163             :     }
    1164             :     else{
    1165           0 :         * accumulator += (* unweightedValue) * weight;
    1166             :     }
    1167           0 : }
    1168             : 
    1169             : 
    1170             : std::pair<Bool, Vector<Double> >
    1171           0 : VbAvg::accumulateCubeData (MsRow * rowInput, MsRowAvg * rowAveraged)
    1172             : {
    1173             :     // Accumulate the sums needed for averaging of cube data (e.g., visibility).
    1174             : 
    1175           0 :     const Matrix<Bool> & inputFlags = rowInput->flags ();
    1176           0 :     Matrix<Bool> & averagedFlags = rowAveraged->flagsMutable ();
    1177           0 :     Matrix<Int>  counts = rowAveraged->counts ();
    1178           0 :     Vector<Bool>  correlationFlagged = rowAveraged->correlationFlagsMutable ();
    1179             : 
    1180           0 :     AccumulationParameters accumulationParameters (rowInput, rowAveraged, doing_p);
    1181             :         // is a member variable to reduce memory allocations (jhj)
    1182             : 
    1183           0 :     IPosition shape = inputFlags.shape();
    1184           0 :     const Int nChannels = shape (1);
    1185           0 :     const Int nCorrelations = shape (0);
    1186             : 
    1187           0 :     Bool rowFlagged = true;  // true if all correlations and all channels flagged
    1188             : 
    1189           0 :     for (Int channel = 0; channel < nChannels; channel ++){
    1190             : 
    1191           0 :         for (Int correlation = 0; correlation < nCorrelations; correlation ++){
    1192             : 
    1193             :             // Based on the current flag state of the accumulation and the current flag
    1194             :             // state of the correlation,channel, accumulate the data (or not).  Accumulate
    1195             :             // flagged data until the first unflagged datum appears.  Then restart the
    1196             :             // accumulation with that datum.
    1197             : 
    1198           0 :             Bool inputFlagged = inputFlags (correlation, channel);
    1199           0 :             if (rowFlagged && ! inputFlagged){
    1200           0 :                 rowFlagged = false;
    1201             :             }
    1202             :             //rowFlagged = rowFlagged && inputFlagged;
    1203           0 :             Bool accumulatorFlagged = averagedFlags (correlation, channel);
    1204             : 
    1205           0 :             if (! accumulatorFlagged && inputFlagged){
    1206           0 :                 accumulationParameters.incrementCubePointers();
    1207           0 :                 continue;// good accumulation, bad data so toss it.
    1208             :             }
    1209             : 
    1210             :             // If changing from flagged to unflagged for this cube element, reset the
    1211             :             // accumulation count to 1; otherwise increment the count.
    1212             : 
    1213           0 :             Bool flagChange = (accumulatorFlagged && ! inputFlagged);
    1214           0 :             Bool zeroAccumulation = flagChange || counts (correlation, channel) == 0;
    1215             : 
    1216           0 :             if (flagChange){
    1217           0 :                 averagedFlags (correlation, channel) = false;
    1218             :             }
    1219             : 
    1220           0 :             if (zeroAccumulation){
    1221           0 :                 counts (correlation, channel) = 1;
    1222             :             }
    1223             :             else{
    1224           0 :                 counts (correlation, channel) += 1;
    1225             :             }
    1226             : 
    1227             :             // Accumulate the sum for each cube element
    1228             : 
    1229           0 :             accumulateElementForCubes (accumulationParameters,
    1230             :                                        zeroAccumulation); // zeroes out accumulation
    1231             : 
    1232           0 :             accumulationParameters.incrementCubePointers();
    1233             : 
    1234             :             // Update correlation Flag
    1235             : 
    1236           0 :             if (correlationFlagged (correlation) && ! inputFlagged){
    1237           0 :                 correlationFlagged (correlation) = false;
    1238             :             }
    1239             :         }
    1240             :     }
    1241             : 
    1242           0 :     Vector<Double> adjustedWeight = Vector<Double> (nCorrelations, 1);
    1243           0 :     if (doing_p.correctedData_p)
    1244             :     {
    1245           0 :         for (Int correlation = 0; correlation < nCorrelations; correlation ++)
    1246             :         {
    1247           0 :                 adjustedWeight(correlation) = rowInput->weight(correlation);
    1248             :         }
    1249             :     }
    1250           0 :     else if (doing_p.observedData_p || doing_p.floatData_p)
    1251             :     {
    1252           0 :         for (Int correlation = 0; correlation < nCorrelations; correlation ++)
    1253             :         {
    1254           0 :                 adjustedWeight(correlation) = AveragingTvi2::sigmaToWeight(rowInput->sigma (correlation));
    1255             :         }
    1256             :     }
    1257             : 
    1258           0 :     return std::make_pair (rowFlagged, adjustedWeight);
    1259           0 : }
    1260             : 
    1261             : 
    1262             : inline void
    1263           0 : VbAvg::accumulateElementForCubes (AccumulationParameters & accumulationParameters,
    1264             :                                   Bool zeroAccumulation)
    1265             : {
    1266             : 
    1267             :         // NOTE: The channelized flag check comes from the calling context (continue statement)
    1268           0 :         float weightCorrected = 1.0f;
    1269           0 :         float weightObserved = 1.0f;
    1270           0 :         const float One = 1.0f;
    1271             : 
    1272           0 :         if (doing_p.correctedData_p)
    1273             :         {
    1274             :                 // The weight corresponding to CORRECTED_DATA is that stored in WEIGHT
    1275           0 :                 weightCorrected = * accumulationParameters.weightSpectrumIn ();
    1276             : 
    1277             : 
    1278             :                 // Accumulate weighted average contribution (normalization will come at the end)
    1279           0 :                 accumulateElementForCube (      accumulationParameters.correctedIn (),
    1280             :                                                                         weightCorrected, zeroAccumulation,
    1281             :                                                                         accumulationParameters.correctedOut ());
    1282             : 
    1283             :                 // The weight resulting from weighted average is the sum of the weights
    1284           0 :                 accumulateElementForCube (      & weightCorrected,
    1285             :                                                                         One, zeroAccumulation,
    1286             :                                                                         accumulationParameters.weightSpectrumOut ());
    1287             :         }
    1288             : 
    1289           0 :         if (doing_p.observedData_p)
    1290             :         {
    1291             :                 // The weight corresponding to DATA is that derived from the rms stored in SIGMA
    1292             :                 // This has to
    1293           0 :                 weightObserved = AveragingTvi2::sigmaToWeight(* accumulationParameters.sigmaSpectrumIn ());
    1294             : 
    1295             :                 // Accumulate weighted average contribution (normalization will come at the end)
    1296             : 
    1297           0 :                 accumulateElementForCube (      accumulationParameters.observedIn (),
    1298             :                                                                         weightObserved, zeroAccumulation,
    1299             :                                                                         accumulationParameters.observedOut ());
    1300             : 
    1301           0 :                 if (not doing_p.correctedData_p)
    1302             :                 {
    1303             :                         // The weight resulting from weighted average is the sum of the weights
    1304           0 :                         accumulateElementForCube (      & weightObserved,
    1305             :                                                                                 One, zeroAccumulation,
    1306             :                                                                                 accumulationParameters.weightSpectrumOut ());
    1307             :                 }
    1308             : 
    1309             :         // This will always create a sigma spectrum column which is not empty.
    1310             :         // This is useful in particular if not doing_p.correctedData_p but doing_p.modelData_p,
    1311             :         // so that modelData can be properly divided by sigmaSpectrumOut in finalizeCubeData
    1312             :         // We store the accumulated weight in sigmaSpectrumOut pending of
    1313             :         // - normalization
    1314             :         // - SIGMA = 1/sqrt(WEIGHT) in-place transformation
    1315           0 :         accumulateElementForCube (&weightObserved,
    1316             :                                   One, zeroAccumulation,
    1317             :                                   accumulationParameters.sigmaSpectrumOut ());
    1318             : 
    1319             :         }
    1320             : 
    1321             :         // For model data is less clear what to do, what in order to convert to
    1322             :         // split we use WEIGHT if averaging CORRECTED_DATA and SIGMA if avg. DATA.
    1323             :         // Finally we use WEIGHT by default when averaging MODEL_DATA only
    1324           0 :         if (doing_p.modelData_p)
    1325             :         {
    1326           0 :                 if (doing_p.correctedData_p)
    1327             :                 {
    1328           0 :                         accumulateElementForCube (      accumulationParameters.modelIn (),
    1329             :                                                                                 weightCorrected, zeroAccumulation,
    1330             :                                                                                 accumulationParameters.modelOut ());
    1331             :                 }
    1332           0 :                 else if (doing_p.observedData_p)
    1333             :                 {
    1334           0 :                         accumulateElementForCube (      accumulationParameters.modelIn (),
    1335             :                                                                                 weightObserved, zeroAccumulation,
    1336             :                                                                                 accumulationParameters.modelOut ());
    1337             :                 }
    1338             :                 else
    1339             :                 {
    1340           0 :                         accumulateElementForCube (      accumulationParameters.modelIn (),
    1341             :                                                                                 One, zeroAccumulation,
    1342             :                                                                                 accumulationParameters.modelOut ());
    1343             : 
    1344             :                         // When doing MODEL_DATA only the accumulated weight spectrum should just represent counts
    1345           0 :                         accumulateElementForCube (      & One,
    1346             :                                                                                 1.0f, zeroAccumulation,
    1347             :                                                                                 accumulationParameters.weightSpectrumOut ());
    1348             :                 }
    1349             :         }
    1350             : 
    1351           0 :         if (doing_p.floatData_p)
    1352             :         {
    1353             : 
    1354             :                 // The weight corresponding to FLOAT_DATA is that derived from the rms stored in SIGMA
    1355           0 :                 weightObserved = AveragingTvi2::sigmaToWeight(* accumulationParameters.sigmaSpectrumIn ());
    1356             : 
    1357             :                 // Accumulate weighted average contribution (normalization will come at the end)
    1358           0 :                 accumulateElementForCube (      accumulationParameters.floatDataIn (),
    1359             :                                                                         weightObserved, zeroAccumulation,
    1360             :                                                                         accumulationParameters.floatDataOut ());
    1361             : 
    1362             :                 // The weight resulting from weighted average is the sum of the weights
    1363           0 :                 accumulateElementForCube (      & weightObserved,
    1364             :                                                                         1.0f, zeroAccumulation,
    1365             :                                                                         accumulationParameters.weightSpectrumOut ());
    1366             :         }
    1367             : 
    1368           0 :         return;
    1369             : }
    1370             : 
    1371             : 
    1372             : 
    1373             : template <typename T>
    1374             : T
    1375           0 : VbAvg::accumulateRowDatum (const T & averagedValue, const T & inputValue, Bool resetAverage)
    1376             : {
    1377           0 :     if (resetAverage){
    1378           0 :         return inputValue;
    1379             :     }
    1380             :     else{
    1381           0 :         return inputValue + averagedValue;
    1382             :     }
    1383             : }
    1384             : 
    1385             : void
    1386           0 : VbAvg::accumulateRowData (MsRow * rowInput, MsRowAvg * rowAveraged,
    1387             :                           Double adjustedWeight, Bool rowFlagged)
    1388             : {
    1389             : 
    1390             :     // Grab working copies of the values to be accumulated.
    1391             : 
    1392           0 :     Bool accumulatorRowFlagged = rowAveraged->isRowFlagged();
    1393           0 :     Bool flagChange = accumulatorRowFlagged != rowFlagged || // actual change
    1394           0 :                       rowAveraged->countsBaseline() == 0; // first time
    1395             : 
    1396           0 :     if (! accumulatorRowFlagged && rowFlagged){
    1397             :         // good accumulation, bad data --> skip it
    1398             :     }
    1399             :     else{
    1400             : 
    1401             :         // Update the row's accumulations; if the flagChanged then zero out the
    1402             :         // previous (flagged) accumulation first.
    1403             : 
    1404           0 :         rowAveraged->setCountsBaseline (accumulateRowDatum (rowAveraged->countsBaseline(),
    1405           0 :                                                             1,
    1406             :                                                             flagChange));
    1407             : 
    1408             :         // The WEIGHT column is handled under accumulateCubeData because of the
    1409             :         // interrelationship between weight and weightSpectrum.  The SIGMA column is
    1410             :         // handled in finalizeBaseline for similar reasons.
    1411             : 
    1412           0 :         accumulatorRowFlagged = accumulatorRowFlagged && rowFlagged;
    1413           0 :         rowAveraged->setRowFlag (accumulatorRowFlagged);
    1414             : 
    1415           0 :         rowAveraged->setExposure (accumulateRowDatum (rowAveraged->exposure(),
    1416           0 :                                                       rowInput->exposure (),
    1417             :                                                       flagChange));
    1418             : 
    1419             :         // While accumulating flagged values, the weights will be zero, so accumulate
    1420             :         // an arithmetic average until the accumulator becomes unflagged.
    1421             : 
    1422             :         Double weightToUse;
    1423             : 
    1424           0 :         if (accumulatorRowFlagged){
    1425           0 :             weightToUse = 1;
    1426             :         }
    1427             :         else{
    1428           0 :             weightToUse = adjustedWeight;
    1429             :         }
    1430             : 
    1431           0 :         if (flagChange){
    1432           0 :             rowAveraged->setNormalizationFactor(0.0);
    1433             :         }
    1434             : 
    1435           0 :         rowAveraged->accumulateNormalizationFactor(weightToUse);
    1436             : 
    1437           0 :         Double weightedTC = (rowInput->timeCentroid() - rowAveraged->timeFirst()) * weightToUse;
    1438           0 :         rowAveraged->setTimeCentroid (accumulateRowDatum (rowAveraged->timeCentroid(),
    1439             :                                                           weightedTC,
    1440             :                                                           flagChange));
    1441             : 
    1442           0 :         Vector<Double> weightedUvw = rowInput->uvw() * weightToUse;
    1443             : 
    1444           0 :         rowAveraged->setUvw (accumulateRowDatum (rowAveraged->uvw (),
    1445             :                                                  weightedUvw,
    1446             :                                                  flagChange));
    1447             : 
    1448             :         // Capture a couple pieces of data
    1449             : 
    1450           0 :         rowAveraged->setTimeLast (rowInput->time());
    1451             : 
    1452           0 :         rowAveraged->setIntervalLast (rowInput->interval());
    1453           0 :     }
    1454           0 : }
    1455             : 
    1456             : //Vector<Float>
    1457             : //VbAvg::adjustWeightForFlags (MsRow * rowInput)
    1458             : //{
    1459             : //    Matrix<Bool> flags = rowInput->flags();
    1460             : //    Vector<Float> adjustedWeight = rowInput->weight();
    1461             : //
    1462             : //    for (Int correlation = 0; correlation < nCorrelations(); correlation++){
    1463             : //
    1464             : //        // Sum up the number of good channels in this correlation
    1465             : //
    1466             : //        Int sum = 0;
    1467             : //
    1468             : //        for (Int channel = 0; channel < nChannels(); channel ++){
    1469             : //
    1470             : //            if (! flags (correlation, channel)){
    1471             : //
    1472             : //                sum ++;
    1473             : //            }
    1474             : //        }
    1475             : //
    1476             : //        // Adjust the weight by multiplying by the fraction of good channels.
    1477             : //
    1478             : //        Float factor = ((float) sum) / nChannels();
    1479             : //        adjustedWeight [correlation] *= factor;
    1480             : //    }
    1481             : //
    1482             : //    return adjustedWeight;
    1483             : //}
    1484             : 
    1485             : const Cube<Int> &
    1486           0 : VbAvg::counts () const
    1487             : {
    1488           0 :     return counts_p;
    1489             : }
    1490             : 
    1491             : 
    1492             : void
    1493           0 : VbAvg::copyIdValues (MsRow * rowInput, MsRowAvg * rowAveraged)
    1494             : {
    1495           0 :     rowAveraged->setAntenna1 (rowInput->antenna1());
    1496           0 :     rowAveraged->setAntenna2 (rowInput->antenna2());
    1497           0 :     rowAveraged->setArrayId (rowInput->arrayId());
    1498           0 :     rowAveraged->setDataDescriptionId (rowInput->dataDescriptionId());
    1499           0 :     rowAveraged->setFeed1 (rowInput->feed1());
    1500           0 :     rowAveraged->setFeed2 (rowInput->feed2());
    1501           0 :     rowAveraged->setFieldId (rowInput->fieldId());
    1502           0 :     rowAveraged->setProcessorId (rowInput->processorId());
    1503           0 :     rowAveraged->setScanNumber (rowInput->scanNumber());
    1504           0 :     rowAveraged->setSpectralWindow (rowInput->spectralWindow());
    1505           0 :     rowAveraged->setObservationId (rowInput->observationId());
    1506           0 :     rowAveraged->setStateId (rowInput->stateId());
    1507           0 : }
    1508             : 
    1509             : void
    1510           0 : VbAvg::copyIdValue (Int inputId, Int & averagedId)
    1511             : {
    1512           0 :     if (averagedId < 0){
    1513           0 :         averagedId = inputId;
    1514             :     }
    1515           0 : }
    1516             : 
    1517             : Bool
    1518           0 : VbAvg::empty () const
    1519             : {
    1520           0 :     return empty_p;
    1521             : }
    1522             : 
    1523             : Int
    1524           0 : VbAvg::getBaselineIndex (const MsRow * msRow) const
    1525             : {
    1526             :     // Lookup the baseline index using the prebuilt lookup table.
    1527             :     //
    1528             :     // The baseline index is the index in the sequence
    1529             :     // {(0,0),(1,0),(1,1),(2,0),(2,1),(2,2), ...} (i.e., the index in a
    1530             :     // 1-d representation of the lower right half of the square matrix of size
    1531             :     // nAntennas).
    1532             :     //
    1533             :     // This handles the case where the baseline ordering in the input VB might
    1534             :     // shift from VB to VB.
    1535             : 
    1536           0 :     const Int antenna1 = msRow->antenna1 ();
    1537           0 :     const Int antenna2 = msRow->antenna2 ();
    1538           0 :     const Int spw = msRow->spectralWindow ();
    1539             : 
    1540           0 :     const Int index = baselineIndex_p (antenna1, antenna2, spw);
    1541             : 
    1542           0 :     return index;
    1543             : }
    1544             : 
    1545             : Int
    1546           0 : VbAvg::getBaselineIndex (Int antenna1, Int antenna2, Int spw) const
    1547             : {
    1548           0 :     const Int index = baselineIndex_p (antenna1, antenna2, spw);
    1549             : 
    1550           0 :     return index;
    1551             : }
    1552             : 
    1553             : void
    1554           0 : VbAvg::finalizeAverages ()
    1555             : {
    1556           0 :     if (empty_p){
    1557           0 :         return; // nothing to finalize
    1558             :     }
    1559             : 
    1560           0 :     MsRowAvg * msRowAvg = getRowMutable (0);
    1561             : 
    1562           0 :     for (Int baseline = 0; baseline < nBaselines(); baseline ++){
    1563             : 
    1564           0 :         msRowAvg->changeRow (baseline);
    1565             : 
    1566           0 :         if (msRowAvg->baselinePresent()){
    1567           0 :             finalizeBaseline (msRowAvg);
    1568             :         }
    1569             : 
    1570             :     }
    1571             : 
    1572           0 :     delete msRowAvg;
    1573             : 
    1574           0 :     empty_p = true;
    1575             : }
    1576             : 
    1577             : void
    1578           0 : VbAvg::finalizeBaseline (MsRowAvg * msRowAvg)
    1579             : {
    1580             :     // Software is no longer supposed to rely on the row flag.
    1581             :     // Setting it to false insures that legacy software will
    1582             :     // have to look at the flag cubes.
    1583             : 
    1584           0 :     msRowAvg->setRowFlag(false);
    1585             : 
    1586           0 :     finalizeCubeData (msRowAvg);
    1587             : 
    1588           0 :     finalizeRowData (msRowAvg);
    1589             : 
    1590           0 :     transferBaseline (msRowAvg);
    1591           0 : }
    1592             : 
    1593             : // Functor to divide variables of possibly different types.
    1594             : // This is unlike std::divides which requires equal types.
    1595             : template <typename L, typename R=L, typename RES=L>
    1596             : struct DividesNonZero : public std::binary_function<L,R,RES>
    1597             : {
    1598           0 :   RES operator() (const L& x, const R& y) const
    1599             :   {
    1600           0 :     { return y > 0? RES(x)/y : RES(x); }
    1601             :   }
    1602             : };
    1603             : 
    1604             : 
    1605             : void
    1606           0 : VbAvg::finalizeCubeData (MsRowAvg * msRowAvg)
    1607             : {
    1608             :     // Divide each of the data cubes in use by the sum of the appropriate weights.
    1609             : 
    1610             :     typedef DividesNonZero <Complex, Float, Complex> DivideOp;
    1611             :     DivideOp op;
    1612             : 
    1613           0 :     if (doing_p.correctedData_p)
    1614             :     {
    1615           0 :         Matrix<Complex> corrected = msRowAvg->correctedMutable();
    1616           0 :         arrayTransformInPlace<Complex, Float, DivideOp > (corrected,msRowAvg->weightSpectrum (), op);
    1617           0 :     }
    1618             : 
    1619           0 :     if (doing_p.observedData_p)
    1620             :     {
    1621           0 :         Matrix<Complex> observed = msRowAvg->observedMutable();
    1622           0 :         if (not doing_p.correctedData_p)
    1623           0 :             arrayTransformInPlace<Complex, Float, DivideOp > (observed,msRowAvg->weightSpectrum (), op);
    1624             :         else
    1625           0 :             arrayTransformInPlace<Complex, Float, DivideOp > (observed,msRowAvg->sigmaSpectrum (), op);
    1626           0 :     }
    1627             : 
    1628           0 :     if (doing_p.modelData_p)
    1629             :     {
    1630           0 :         Matrix<Complex> model = msRowAvg->modelMutable();
    1631             : 
    1632           0 :         if (doing_p.correctedData_p)
    1633           0 :             arrayTransformInPlace<Complex, Float, DivideOp > (model,msRowAvg->weightSpectrum (), op);
    1634           0 :         else if (doing_p.observedData_p)
    1635           0 :             arrayTransformInPlace<Complex, Float, DivideOp > (model,msRowAvg->sigmaSpectrum (), op);
    1636             :         else
    1637           0 :             arrayTransformInPlace<Complex, Int, DivideOp > (model,msRowAvg->counts (), op);
    1638           0 :     }
    1639             : 
    1640           0 :     if (doing_p.floatData_p)
    1641             :     {
    1642             :         typedef Divides <Float, Float, Float> DivideOpFloat;
    1643             :         DivideOpFloat opFloat;
    1644             : 
    1645           0 :         Matrix<Float> visCubeFloat = msRowAvg->singleDishDataMutable();
    1646           0 :         arrayTransformInPlace<Float, Float, DivideOpFloat > (visCubeFloat,msRowAvg->weightSpectrum (), opFloat);
    1647           0 :     }
    1648             : 
    1649             : 
    1650           0 :     return;
    1651             : }
    1652             : 
    1653             : void
    1654           0 : VbAvg::finalizeRowData (MsRowAvg * msRow)
    1655             : {
    1656           0 :     Int n = msRow->countsBaseline ();
    1657             : 
    1658             :     // Obtain row-level WEIGHT by calculating the mean of WEIGHT_SPECTRUM
    1659             :     // msRow->setWeight(partialMedians(msRow->weightSpectrum(),IPosition(1,1),true));
    1660           0 :     msRow->setWeight(AveragingTvi2::average(msRow->weightSpectrum(),msRow->flags()));
    1661             : 
    1662             :     // If doing both DATA and CORRECTED_DATA then SIGMA_SPECTRUM contains the weight
    1663             :     // (not sigma) accumulation for DATA, and we have to derive SIGMA from it
    1664           0 :     if (doing_p.correctedData_p and doing_p.observedData_p)
    1665             :     {
    1666             :         // jagonzal: SIGMA is not derived from the mean of SIGMA_SPECTRUM but from the mean
    1667             :         // of the WEIGHT format of SIGMA_SPECTRUM turned into SIGMA by using 1/pow(weight,2)
    1668             :         // Vector<Float> weight = partialMedians(msRow->sigmaSpectrum(),IPosition(1,1),true);
    1669           0 :         Vector<Float> weight = AveragingTvi2::average(msRow->sigmaSpectrum(),msRow->flags());
    1670           0 :         arrayTransformInPlace (weight, AveragingTvi2::weightToSigma);
    1671           0 :         msRow->setSigma (weight);
    1672             : 
    1673             :         // Now convert the DATA weight accumulation stored in sigmaSpectrum into the real SIGMA_SPECTRUM
    1674             :         // TODO: This should happen only if we are writing out SIGMA_SPECTRUM but
    1675             :         //       multiple column operation is rare and might be forbidden in the future
    1676           0 :         Matrix<Float> sigmaSpectrun = msRow->sigmaSpectrum(); // Reference copy
    1677           0 :         arrayTransformInPlace (sigmaSpectrun, AveragingTvi2::weightToSigma);
    1678           0 :     }
    1679             :     // Otherwise (doing only DATA/FLOAT_DATA or CORRECTED_DATA) we can derive SIGMA from WEIGHT directly
    1680           0 :     else if ( not doing_p.onlymetadata_p)
    1681             :     {
    1682             :         // jagonzal: SIGMA is not derived from the mean of SIGMA_SPECTRUM
    1683             :         // but from WEIGHT turned into SIGMA by using 1/pow(weight,2)
    1684           0 :         Vector<Float> sigma = msRow->sigma(); // Reference copy
    1685           0 :         sigma = msRow->weight(); // Normal copy (transfer Weight values to Sigma)
    1686           0 :         arrayTransformInPlace (sigma, AveragingTvi2::weightToSigma);
    1687             : 
    1688             :         // Derive SIGMA_SPECTRUM from computed WEIGHT_SPECTRUM
    1689           0 :         Matrix<Float> sigmaSpectrun = msRow->sigmaSpectrum(); // Reference copy
    1690           0 :         sigmaSpectrun = msRow->weightSpectrum(); // Normal copy (transfer WeightSpectrum values to SigmaSpectrum)
    1691           0 :         arrayTransformInPlace (sigmaSpectrun, AveragingTvi2::weightToSigma);
    1692           0 :     }
    1693             : 
    1694             :     // Get the normalization factor for this baseline, containing
    1695             :     // the accumulation of row-level) weight contributions
    1696           0 :     Double weight = msRow->getNormalizationFactor();
    1697             : 
    1698           0 :     if (n != 0){
    1699             : 
    1700           0 :         if (weight == 0){
    1701             : 
    1702             :             // The weights are all zero so compute an arithmetic average
    1703             :             // so that a somewhat value can go into these columns (i.e. rather than NaN).
    1704             : 
    1705           0 :             weight = msRow->countsBaseline();
    1706             :         }
    1707             : 
    1708           0 :         msRow->setTimeCentroid (msRow->timeCentroid() / weight + msRow->timeFirst());
    1709             : 
    1710           0 :         msRow->setUvw (msRow->uvw() / weight);
    1711             : 
    1712             :         // Exposure is a simple sum, not an average so it is already
    1713             :         // done at this point.
    1714             :     }
    1715             : 
    1716             :     // Fill in the time and the interval
    1717             :     //
    1718             :     // The time of a sample is centered over the integration time period.
    1719             :     // Thus find the center of the averaged interval it is necessary to
    1720             :     // slide it back by 1/2 an interval.
    1721             : 
    1722           0 :     Double dT = msRow->timeLast () - msRow->timeFirst();
    1723             : 
    1724           0 :     Double centerOfInterval = msRow->timeFirst () + dT / 2;
    1725             : 
    1726           0 :     msRow->setTime (centerOfInterval);
    1727             : 
    1728           0 :     if (dT != 0){
    1729             : 
    1730             :         // The interval is the center-to-center time + half of the intervals of
    1731             :         // the first and the last sample (if dT == 0 then the interval is
    1732             :         // already the interval of the first sample and is correct)
    1733             : 
    1734           0 :         Double interval = dT + msRow->interval() / 2 + msRow->intervalLast() / 2;
    1735           0 :         msRow->setInterval (interval);
    1736             :     }
    1737           0 : }
    1738             : 
    1739             : void
    1740           0 : VbAvg::finalizeBaselineIfNeeded (MsRow * rowInput, MsRowAvg * rowAveraged, const Subchunk & /*subchunk*/)
    1741             : {
    1742           0 :     if (! rowAveraged->baselinePresent()){
    1743           0 :         return;
    1744             :     }
    1745             : 
    1746             :     // Finalization is needed if either the uvw distance or the time distance between the input
    1747             :     // baseline and the averaged baseline is above the maximum
    1748             : 
    1749           0 :     Bool needed = usingUvwDistance_p;
    1750             : 
    1751           0 :     if (needed) {
    1752           0 :         Double deltaUvw = distanceSquared (rowInput->uvw(), rowAveraged->uvwFirst ());
    1753           0 :         needed = deltaUvw > maxUvwDistanceSquared_p;
    1754             :     }
    1755             : 
    1756           0 :     needed = needed || (rowInput->time() - rowAveraged->timeFirst()) > maxTimeDistance_p;
    1757             : 
    1758           0 :     if (needed){
    1759             : 
    1760             :         // Finalize the data so that the final averaging products and then move them to
    1761             :         // output buffer.
    1762             : 
    1763           0 :         finalizeBaseline (rowAveraged);
    1764             :     }
    1765             : }
    1766             : 
    1767             : MsRowAvg *
    1768           0 : VbAvg::getRow (Int row) const
    1769             : {
    1770           0 :     return new MsRowAvg (row, this);
    1771             : }
    1772             : 
    1773             : MsRowAvg *
    1774           0 : VbAvg::getRowMutable (Int row)
    1775             : {
    1776           0 :     return new MsRowAvg (row, this);
    1777             : }
    1778             : 
    1779             : void
    1780           0 : VbAvg::initializeBaseline (MsRow * rowInput, MsRowAvg * rowAveraged,
    1781             :                            const Subchunk & /*subchunk*/)
    1782             : {
    1783           0 :     copyIdValues (rowInput, rowAveraged);
    1784             : 
    1785             :     // Size and zero out the counters
    1786             : 
    1787           0 :     rowAveraged->setInterval (rowInput->interval()); // capture first one
    1788           0 :     rowAveraged->setTimeFirst (rowInput->time());
    1789           0 :     rowAveraged->setTimeLast (rowInput->time());
    1790           0 :     rowAveraged->uvwFirst () = rowInput->uvw ();
    1791             : 
    1792           0 :     rowAveraged->setCountsBaseline (0);
    1793             : 
    1794           0 :     IPosition shape = rowInput->flags().shape();
    1795           0 :     Int nCorrelations = shape (0);
    1796           0 :     Int nChannels = shape (1);
    1797             : 
    1798           0 :     rowAveraged->setCounts (zeroInt_p.getMatrix (nCorrelations, nChannels, 0));
    1799           0 :     rowAveraged->setWeight (Vector<Float> (nCorrelations, 0));
    1800           0 :     rowAveraged->setTimeCentroid (0.0);
    1801             : 
    1802           0 :     if (doing_p.weightSpectrumOut_p){
    1803           0 :         rowAveraged->setWeightSpectrum (zeroFloat_p.getMatrix (nCorrelations, nChannels, 0));
    1804             :     }
    1805             : 
    1806           0 :     if (doing_p.sigmaSpectrumOut_p){
    1807           0 :         rowAveraged->setSigmaSpectrum (zeroFloat_p.getMatrix (nCorrelations, nChannels, 0));
    1808             :     }
    1809             : 
    1810             : //    VisBufferComponents2 exclusions =
    1811             : //        VisBufferComponents2::these(VisibilityObserved, VisibilityCorrected,
    1812             : //                                    VisibilityModel, CorrType, JonesC, Unknown);
    1813             : //
    1814             : //    cacheResizeAndZero(exclusions);
    1815             : 
    1816             :     // Flag everything to start with
    1817             : 
    1818           0 :     rowAveraged->setRowFlag (true); // only for use during row-value accumulation
    1819           0 :     rowAveraged->setFlags(trueBool_p.getMatrix (nCorrelations, nChannels, true));
    1820           0 :     rowAveraged->correlationFlagsMutable() = Vector<Bool> (nCorrelations, true);
    1821             : 
    1822           0 :     rowAveraged->setBaselinePresent(true);
    1823             : 
    1824           0 :     rowAveraged->setNormalizationFactor(0.0);
    1825           0 : }
    1826             : 
    1827             : 
    1828             : Bool
    1829           0 : VbAvg::isComplete () const
    1830             : {
    1831           0 :     return complete_p;
    1832             : }
    1833             : 
    1834             : Bool
    1835           0 : VbAvg::isUsingUvwDistance () const
    1836             : {
    1837           0 :     return usingUvwDistance_p;
    1838             : }
    1839             : 
    1840             : 
    1841             : //void
    1842             : //VbAvg::markEmpty ()
    1843             : //{
    1844             : //    empty_p = true;
    1845             : //    complete_p = false;
    1846             : //}
    1847             : 
    1848             : Int
    1849           0 : VbAvg::nBaselines () const
    1850             : {
    1851           0 :     return countsBaseline_p.nelements();
    1852             : }
    1853             : 
    1854             : Int
    1855           0 : VbAvg::nSpectralWindowsInBuffer () const
    1856             : {
    1857           0 :     const Vector<Int> & windows = spectralWindows();
    1858           0 :     set <Int> s;
    1859             : 
    1860           0 :     for (uInt i = 0; i < windows.nelements(); i ++){
    1861           0 :         s.insert (windows(i));
    1862             :     }
    1863             : 
    1864           0 :     return (Int) s.size();
    1865             : 
    1866           0 : }
    1867             : 
    1868             : 
    1869             : void
    1870           0 : VbAvg::captureIterationInfo (VisBufferImpl2 * dstVb, const VisBuffer2 * srcVb,
    1871             :                              const Subchunk & subchunk)
    1872             : {
    1873           0 :     dstVb->setIterationInfo (srcVb->msId(),
    1874           0 :                              srcVb->msName(),
    1875           0 :                              srcVb->isNewMs(),
    1876           0 :                              srcVb->isNewArrayId(),
    1877           0 :                              srcVb->isNewFieldId(),
    1878           0 :                              srcVb->isNewSpectralWindow(),
    1879             :                              subchunk,
    1880           0 :                              srcVb->getCorrelationTypes (),
    1881           0 :                              srcVb->getCorrelationTypesDefined(),
    1882           0 :                              srcVb->getCorrelationTypesSelected(),
    1883           0 :                              CountedPtr <WeightScaling> ( ));
    1884             : 
    1885             :     // Request info from the VB which will cause it to be filled
    1886             :     // into cache from the input VII at its current position.
    1887             : 
    1888           0 :     dstVb->setRekeyable(true);
    1889           0 :     dstVb->setShape(srcVb->nCorrelations(), srcVb->nChannels(), nBaselines(), false);
    1890             :         // Do not clear the cache since we're resuing the storage
    1891             : 
    1892           0 :     dstVb->phaseCenter();
    1893           0 :     dstVb->nAntennas();
    1894           0 :     dstVb->correlationTypes();
    1895           0 :     dstVb->polarizationFrame();
    1896           0 :     dstVb->polarizationId();
    1897           0 : }
    1898             : 
    1899             : //void
    1900             : //VbAvg::prepareForFirstData (const VisBuffer2 * vb, const Subchunk & subchunk)
    1901             : //{
    1902             : //    startTime_p = vb->time() (0);
    1903             : //    sampleInterval_p = vb->timeInterval() (0);
    1904             : //
    1905             : //    Int nAntennas = vb->nAntennas();
    1906             : //    Int nSpw = vb->getVi()->nSpectralWindows();
    1907             : //    Int nBaselines = ((nAntennas * (nAntennas + 1)) / 2) * nSpw;
    1908             : //
    1909             : //    // Size and zero out the counters
    1910             : //
    1911             : //    timeFirst_p = Vector<Double> (nBaselines, vb->time() (0));
    1912             : //    timeLast_p = Vector<Double> (nBaselines, vb->time() (0));
    1913             : //    uvwFirst_p = Vector<Double> (nBaselines, vb->uvw().column(0));
    1914             : //
    1915             : //    countsBaseline_p = Vector<Int> (nBaselines, 0);
    1916             : //    counts_p = Cube<Int> (vb->nCorrelations(), vb->nChannels(), nBaselines, 0);
    1917             : //    weightSum_p = Cube<Float> (vb->nCorrelations(), vb->nChannels(), nBaselines, 0);
    1918             : //    if (doing_p.sigmaSpectrum_p){
    1919             : //        weightCorrectedSum_p = Cube<Float> (vb->nCorrelations(), vb->nChannels(), nBaselines, 0);
    1920             : //    }
    1921             : //
    1922             : //    baselineIndex_p.configure (nAntennas, nSpw, vb);
    1923             : //
    1924             : //    // Reshape the inherited members from VisBuffer2
    1925             : //
    1926             : //    captureIterationInfo (vb, subchunk);
    1927             : //
    1928             : //    setShape (vb->nCorrelations(), vb->nChannels(), nBaselines, false);
    1929             : //
    1930             : //    VisBufferComponents2 exclusions =
    1931             : //        VisBufferComponents2::these(VisibilityObserved, VisibilityCorrected,
    1932             : //                                    VisibilityModel, CorrType, JonesC, Unknown);
    1933             : //    cacheResizeAndZero(exclusions);
    1934             : //
    1935             : //    prepareIds (vb);
    1936             : //
    1937             : //    // Flag everything to start with
    1938             : //
    1939             : //    setFlagCube (Cube<Bool> (vb->nCorrelations(), vb->nChannels(), nBaselines, true));
    1940             : //    setFlagRow (Vector<Bool> (nBaselines, true));
    1941             : //
    1942             : //    complete_p = false;
    1943             : //}
    1944             : 
    1945             : void
    1946           0 : VbAvg::prepareIds (const VisBuffer2 * vb)
    1947             : {
    1948             :     // Set these row ID values to indicate they are unknown
    1949             : 
    1950           0 :     Vector<Int> minusOne (nBaselines(), -1);
    1951             : 
    1952           0 :     setAntenna1 (minusOne);
    1953           0 :     setAntenna2 (minusOne);
    1954           0 :     setDataDescriptionIds (minusOne);
    1955           0 :     setFeed1 (minusOne);
    1956           0 :     setFeed2 (minusOne);
    1957           0 :     setProcessorId (minusOne);
    1958           0 :     setScan (minusOne);
    1959           0 :     setObservationId (minusOne);
    1960           0 :     setSpectralWindows (minusOne);
    1961           0 :     setStateId (minusOne);
    1962             : 
    1963             :     // Copy the value from the input VB
    1964             : 
    1965           0 :     Vector<Int> tmp (nBaselines(), vb->arrayId()(0));
    1966             : 
    1967           0 :     setArrayId (tmp);
    1968             : 
    1969           0 :     tmp = vb->dataDescriptionIds()(0);
    1970           0 :     setDataDescriptionIds (tmp);
    1971             : 
    1972           0 :     tmp = vb->fieldId()(0);
    1973           0 :     setFieldId (tmp);
    1974           0 : }
    1975             : 
    1976             : //void
    1977             : //VbAvg::removeMissingBaselines ()
    1978             : //{
    1979             : //    // Some baselines may not be present in the portion of the input data
    1980             : //    // that made up this average.  However, this is not known until after
    1981             : //    // all of the data is seen.  Thus at finalization time these bogus
    1982             : //    // baselines should be removed from the average so as not to pass
    1983             : //    // flagged but zero-exposure baselines into the output.
    1984             : //
    1985             : //
    1986             : //    Vector<Int> rowsToDelete (nBaselines());
    1987             : //
    1988             : //    Int nBaselinesDeleted = 0;
    1989             : //
    1990             : //    for (Int baseline = 0; baseline < nBaselines(); baseline ++){
    1991             : //
    1992             : //        if (countsBaseline_p (baseline) == 0){
    1993             : //            rowsToDelete (nBaselinesDeleted) = baseline;
    1994             : //            nBaselinesDeleted ++;
    1995             : //        }
    1996             : //    }
    1997             : //
    1998             : //    rowsToDelete.resize (nBaselinesDeleted, true);
    1999             : //
    2000             : //    deleteRows (rowsToDelete);
    2001             : //}
    2002             : 
    2003             : void
    2004           0 : VbAvg::setBufferToFill(VisBufferImpl2 * vb)
    2005             : {
    2006           0 :     bufferToFill_p = vb;
    2007             : 
    2008             :     // Set flag so that iteration information will be captured into
    2009             :     // this VB from the first input VB.
    2010             : 
    2011           0 :     needIterationInfo_p = true;
    2012           0 : }
    2013             : 
    2014             : void
    2015           0 : VbAvg::setupVbAvg (const VisBuffer2 * vb)
    2016             : {
    2017             :     // Configure the index
    2018             : 
    2019           0 :     Int nAntennas = vb->nAntennas();
    2020             : 
    2021             :     // This is a kluge to allow multiple spectral windows (of the same shape)
    2022             :     // to be combined into a single VB.  This really shouldn't be allowed!!!
    2023             : 
    2024           0 :     set<uInt> spwInVb;
    2025             : 
    2026           0 :     for (rownr_t i = 0; i < vb->nRows(); i++){
    2027           0 :         spwInVb.insert (vb->dataDescriptionIds()(i));
    2028             :     }
    2029             : 
    2030           0 :     uInt nSpwInVb = spwInVb.size();
    2031             : 
    2032           0 :     Int nSpw = averagingVii_p->nSpectralWindows();
    2033             : 
    2034           0 :     baselineIndex_p.configure (nAntennas, nSpw, vb);
    2035             : 
    2036           0 :     Int nBaselines = ((nAntennas * (nAntennas + 1)) / 2)* nSpwInVb;
    2037             : 
    2038           0 :     setShape (vb->nCorrelations(), vb->nChannels(), nBaselines);
    2039             : 
    2040           0 :     setupArrays (vb->nCorrelations(), vb->nChannels(), nBaselines);
    2041             : 
    2042           0 :     baselinePresent_p.resize(nBaselines);
    2043           0 :     baselinePresent_p = False;
    2044             : 
    2045           0 :     normalizationFactor_p.resize(nBaselines);
    2046           0 :     normalizationFactor_p = 0.0;
    2047             : 
    2048           0 :     empty_p = false;
    2049           0 : }
    2050             : 
    2051             : void
    2052           0 : VbAvg::setupArrays (Int nCorrelations, Int nChannels, Int nBaselines)
    2053             : {
    2054             : 
    2055           0 :     setShape (nCorrelations, nChannels, nBaselines);
    2056             : 
    2057             :     // Start out with all of the array-valued components except the
    2058             :     // optional ones.
    2059             : 
    2060             :     VisBufferComponents2 including =
    2061             :         VisBufferComponents2::these ({VisBufferComponent2::Antenna1,
    2062             :                     VisBufferComponent2::Antenna2,
    2063             :                     VisBufferComponent2::ArrayId,
    2064             :                     VisBufferComponent2::CorrType,
    2065             :                     VisBufferComponent2::DataDescriptionIds,
    2066             :                     VisBufferComponent2::Exposure,
    2067             :                     VisBufferComponent2::Feed1,
    2068             :                     VisBufferComponent2::Feed2,
    2069             :                     VisBufferComponent2::FieldId,
    2070             :                     VisBufferComponent2::FlagCategory,
    2071             :                     VisBufferComponent2::FlagCube,
    2072             :                     VisBufferComponent2::FlagRow,
    2073             :                     VisBufferComponent2::ObservationId,
    2074             :                     VisBufferComponent2::ProcessorId,
    2075             :                     VisBufferComponent2::RowIds,
    2076             :                     VisBufferComponent2::Scan,
    2077             :                     VisBufferComponent2::Sigma,
    2078             :                     VisBufferComponent2::SpectralWindows,
    2079             :                     VisBufferComponent2::StateId,
    2080             :                     VisBufferComponent2::Time,
    2081             :                     VisBufferComponent2::TimeCentroid,
    2082             :                     VisBufferComponent2::TimeInterval,
    2083             :                     VisBufferComponent2::Weight,
    2084           0 :                     VisBufferComponent2::Uvw});
    2085             : 
    2086           0 :     if (doing_p.modelData_p){
    2087           0 :         including += VisBufferComponent2::VisibilityCubeModel;
    2088             :     }
    2089             : 
    2090           0 :     if (doing_p.correctedData_p){
    2091           0 :         including += VisBufferComponent2::VisibilityCubeCorrected;
    2092             :     }
    2093             : 
    2094           0 :     if (doing_p.observedData_p){
    2095           0 :         including += VisBufferComponent2::VisibilityCubeObserved;
    2096             :     }
    2097             : 
    2098           0 :     if (doing_p.floatData_p){
    2099           0 :         including += VisBufferComponent2::VisibilityCubeFloat;
    2100             :     }
    2101             : 
    2102           0 :     if (doing_p.weightSpectrumOut_p){
    2103           0 :         including += VisBufferComponent2::WeightSpectrum;
    2104             :     }
    2105             : 
    2106           0 :     if (doing_p.sigmaSpectrumOut_p){
    2107           0 :         including += VisBufferComponent2::SigmaSpectrum;
    2108             :     }
    2109             : 
    2110           0 :     cacheResizeAndZero (including);
    2111             : 
    2112           0 :     correlationFlags_p.reference (Matrix<Bool> (IPosition (2, nCorrelations, nBaselines), true));
    2113           0 :     counts_p.reference (Cube<Int> (IPosition (3, nCorrelations, nChannels, nBaselines), 0));
    2114           0 :     countsBaseline_p.reference (Vector<Int> (nBaselines, 0)); // number of items summed together for each baseline.
    2115           0 :     intervalLast_p.reference (Vector<Double> (nBaselines, 0));
    2116           0 :     timeFirst_p.reference (Vector<Double> (nBaselines, 0));
    2117           0 :     timeLast_p.reference (Vector<Double> (nBaselines, 0));
    2118           0 :     uvwFirst_p.reference (Matrix<Double> (IPosition (2, 3, nBaselines), 0.0));
    2119           0 : }
    2120             : 
    2121             : void
    2122           0 : VbAvg::startChunk (ViImplementation2 * vi)
    2123             : {
    2124           0 :     empty_p = true;
    2125             : 
    2126           0 :     rowIdGenerator_p = 0;
    2127           0 :     row2AvgRow_p.resize(0);
    2128             :     
    2129             :     // See if the new MS has corrected and/or model data columns
    2130             : 
    2131           0 :     doing_p.observedData_p = averagingOptions_p.contains (AveragingOptions::AverageObserved);
    2132           0 :     doing_p.correctedData_p = vi->existsColumn (VisBufferComponent2::VisibilityCubeCorrected) &&
    2133           0 :                               averagingOptions_p.contains (AveragingOptions::AverageCorrected);
    2134           0 :     doing_p.modelData_p = vi->existsColumn (VisBufferComponent2::VisibilityCubeModel) &&
    2135           0 :                           averagingOptions_p.contains (AveragingOptions::AverageModel);
    2136           0 :     doing_p.floatData_p = vi->existsColumn (VisBufferComponent2::VisibilityCubeFloat) &&
    2137           0 :                           averagingOptions_p.contains (AveragingOptions::AverageFloat);
    2138             : 
    2139           0 :     doing_p.weightSpectrumIn_p = doing_p.correctedData_p;
    2140           0 :     doing_p.sigmaSpectrumIn_p = doing_p.observedData_p || doing_p.floatData_p;
    2141           0 :     doing_p.weightSpectrumOut_p = true; // We always use the output WeightSpectrum
    2142           0 :     doing_p.sigmaSpectrumOut_p = true; // We always use the output SigmaSpectrum
    2143             : 
    2144           0 :     if (doing_p.observedData_p or doing_p.correctedData_p or doing_p.modelData_p or doing_p.floatData_p)
    2145             :     {
    2146           0 :         doing_p.onlymetadata_p = false;
    2147             :     }
    2148             : 
    2149             :     // Set up the flags for row copying
    2150             : 
    2151           0 :     optionalComponentsToCopy_p = VisBufferComponents2::none();
    2152             : 
    2153           0 :     if (doing_p.observedData_p){
    2154           0 :         optionalComponentsToCopy_p += VisBufferComponent2::VisibilityCubeObserved;
    2155             :     }
    2156             : 
    2157           0 :     if (doing_p.correctedData_p){
    2158           0 :         optionalComponentsToCopy_p += VisBufferComponent2::VisibilityCubeCorrected;
    2159             :     }
    2160             : 
    2161           0 :     if (doing_p.modelData_p){
    2162           0 :         optionalComponentsToCopy_p += VisBufferComponent2::VisibilityCubeModel;
    2163             :     }
    2164             : 
    2165           0 :     if (doing_p.floatData_p){
    2166           0 :         optionalComponentsToCopy_p += VisBufferComponent2::VisibilityCubeFloat;
    2167             :     }
    2168             : 
    2169           0 :     if (doing_p.weightSpectrumOut_p){
    2170           0 :         optionalComponentsToCopy_p += VisBufferComponent2::WeightSpectrum;
    2171             :     }
    2172             : 
    2173           0 :     if (doing_p.sigmaSpectrumOut_p){
    2174           0 :         optionalComponentsToCopy_p += VisBufferComponent2::SigmaSpectrum;
    2175             :     }
    2176           0 : }
    2177             : 
    2178             : void
    2179           0 : VbAvg::transferBaseline (MsRowAvg * rowAveraged)
    2180             : {
    2181           0 :     rowAveraged->setRowId (rowIdGenerator_p ++);
    2182           0 :     bufferToFill_p->appendRow (rowAveraged, nBaselines(), optionalComponentsToCopy_p);
    2183             : 
    2184           0 :     rowAveraged->setBaselinePresent(false);
    2185           0 : }
    2186             : 
    2187             : 
    2188             : //VbSet::VbSet (const AveragingParameters & averagingParameters)
    2189             : //: averagingInterval_p (averagingParameters.getAveragingInterval ()),
    2190             : //  averagingOptions_p (averagingParameters.getOptions()),
    2191             : //  averagingParameters_p (averagingParameters),
    2192             : //  doingCorrectedData_p (false),
    2193             : //  doingModelData_p (false),
    2194             : //  doingObservedData_p (false),
    2195             : //  doingWeightSpectrum_p (false),
    2196             : //  doingsigmaSpectrum_p (false),
    2197             : //  vbAveragers_p ()
    2198             : //{}
    2199             : 
    2200             : //VbSet::~VbSet ()
    2201             : //{
    2202             : //    flush (true); // allow killing nonempty buffers
    2203             : //}
    2204             : //
    2205             : //void
    2206             : //VbSet::accumulate (const VisBuffer2 * input, const Subchunk & subchunk)
    2207             : //{
    2208             : //    Int ddId = input->dataDescriptionIds()(0);
    2209             : //
    2210             : //    if (! utilj::containsKey (ddId, vbAveragers_p)){ // Need a new averager
    2211             : //        add (ddId);
    2212             : //    }
    2213             : //
    2214             : //    VbAvg * vba = vbAveragers_p [ddId];
    2215             : //
    2216             : //    vba->accumulate (input, subchunk);
    2217             : //}
    2218             : //
    2219             : //VbAvg *
    2220             : //VbSet::add (Int ddId)
    2221             : //{
    2222             : //    VbAvg::Doing doing;
    2223             : //
    2224             : //    doing.correctedData_p = doingCorrectedData_p;
    2225             : //    doing.modelData_p = doingModelData_p;
    2226             : //    doing.observedData_p = doingObservedData_p;
    2227             : //    doing.weightSpectrum_p = doingWeightSpectrum_p;
    2228             : //    doing.sigmaSpectrum_p = doingsigmaSpectrum_p;
    2229             : //
    2230             : //    VbAvg * newAverager =  new VbAvg (doing, averagingParameters_p);
    2231             : //
    2232             : //    vbAveragers_p [ddId] = newAverager;
    2233             : //
    2234             : //    return newAverager;
    2235             : //}
    2236             : //
    2237             : //Bool
    2238             : //VbSet::anyAveragesReady(Int ddid) const
    2239             : //{
    2240             : //    Bool any = false;
    2241             : //
    2242             : //    for (Averagers::const_iterator a = vbAveragers_p.begin();
    2243             : //         a != vbAveragers_p.end();
    2244             : //         a++){
    2245             : //
    2246             : //        if (a->second->isComplete() &&
    2247             : //            (ddid < 0 || ddid == a->second->dataDescriptionIds()(0))){
    2248             : //
    2249             : //            any = true;
    2250             : //            break;
    2251             : //        }
    2252             : //    }
    2253             : //
    2254             : //    return any;
    2255             : //}
    2256             : //
    2257             : //Bool
    2258             : //VbSet::contains (Int ddId) const
    2259             : //{
    2260             : //    return vbAveragers_p.find (ddId) != vbAveragers_p.end();
    2261             : //}
    2262             : //
    2263             : ////void
    2264             : ////VbSet::finalizeAverage (Int ddId)
    2265             : ////{
    2266             : ////    vbAveragers_p [ddId]->finalizeAverage();
    2267             : ////}
    2268             : //
    2269             : //void
    2270             : //VbSet::finalizeAverages ()
    2271             : //{
    2272             : ////    for (Averagers::iterator a = vbAveragers_p.begin();
    2273             : ////         a != vbAveragers_p.end();
    2274             : ////         a ++){
    2275             : ////
    2276             : ////        finalizeAverage (a->first);
    2277             : ////    }
    2278             : //}
    2279             : //
    2280             : //void
    2281             : //VbSet::flush (Bool okIfNonempty, ViImplementation2 * vi)
    2282             : //{
    2283             : //    // Get rid of all of the averagers.  This is done at
    2284             : //    // destruction time and when a sweeping into a new MS.
    2285             : //
    2286             : //    for (Averagers::const_iterator a = vbAveragers_p.begin();
    2287             : //         a != vbAveragers_p.end();
    2288             : //         a ++){
    2289             : //
    2290             : //        Assert (okIfNonempty || (a->second)->empty());
    2291             : //            // should have been emptied before calling this
    2292             : //
    2293             : //        delete a->second;
    2294             : //    }
    2295             : //
    2296             : //    vbAveragers_p.clear();
    2297             : //
    2298             : //    seeIfCubeColumnsExist (vi);
    2299             : //}
    2300             : //
    2301             : //void
    2302             : //VbSet::seeIfCubeColumnsExist (ViImplementation2 * vi)
    2303             : //{
    2304             : //    if (vi != 0){
    2305             : //
    2306             : //        // See if the new MS has corrected and/or model data columns
    2307             : //
    2308             : //        doingObservedData_p = averagingOptions_p.contains (AveragingOptions::AverageObserved);
    2309             : //        doingCorrectedData_p = vi->existsColumn (VisibilityCubeCorrected) &&
    2310             : //                               averagingOptions_p.contains (AveragingOptions::AverageCorrected);
    2311             : //        doingModelData_p = vi->existsColumn (VisibilityCubeModel) &&
    2312             : //                               averagingOptions_p.contains (AveragingOptions::AverageModel);
    2313             : //        doingWeightSpectrum_p = vi->existsColumn (WeightSpectrum);
    2314             : //
    2315             : //        // If the use of corrected weights were specified for one of the averages, it's an
    2316             : //        // error if the column does not exist.  Also set the doing flag for corrected weights
    2317             : //        // if it's being used in some way.
    2318             : //
    2319             : //        Bool needCorrectedWeights =
    2320             : //            averagingOptions_p.contains (AveragingOptions::ModelUseCorrectedWeights) ||
    2321             : //            averagingOptions_p.contains (AveragingOptions::CorrectedUseCorrectedWeights);
    2322             : //
    2323             : //        Bool correctedWeightsMissing = needCorrectedWeights &&
    2324             : //                                       ! vi->existsColumn (sigmaSpectrum);
    2325             : //
    2326             : //        ThrowIf (correctedWeightsMissing,
    2327             : //                 "Corrected_weight_spectrum not present but required by provided averaging options");
    2328             : //
    2329             : //        doingsigmaSpectrum_p = needCorrectedWeights;
    2330             : //    }
    2331             : //}
    2332             : //
    2333             : //Int
    2334             : //VbSet::getFirstReadyDdid () const
    2335             : //{
    2336             : //    for (Averagers::const_iterator a = vbAveragers_p.begin();
    2337             : //         a != vbAveragers_p.end();
    2338             : //         a ++){
    2339             : //
    2340             : //        if (a->second->isComplete()){
    2341             : //            return a->first;
    2342             : //        }
    2343             : //    }
    2344             : //
    2345             : //    Assert (false); // ought to have been one that's ready
    2346             : //
    2347             : //    return -1; // shouldn't be called
    2348             : //}
    2349             : //
    2350             : ////void
    2351             : ////VbSet::transferAverage (Int ddId, VisBuffer2 * vb)
    2352             : ////{
    2353             : ////    Assert (utilj::containsKey (ddId, vbAveragers_p));
    2354             : ////
    2355             : ////    VbAvg * vba = vbAveragers_p [ddId];
    2356             : ////
    2357             : ////    Assert (vba != 0 && ! vba->empty ());
    2358             : ////
    2359             : ////    // Copy< the completed average into the provided VisBuffer, but
    2360             : ////    // first reshape the VB if it's shape is different.
    2361             : ////
    2362             : ////    vba->transferAverage (vb);
    2363             : ////
    2364             : ////}
    2365             : //
    2366             : //void
    2367             : //VbSet::zero ()
    2368             : //{
    2369             : //    for (Averagers::const_iterator a = vbAveragers_p.begin();
    2370             : //         a != vbAveragers_p.end();
    2371             : //         a ++){
    2372             : //
    2373             : //        a->second->markEmpty();
    2374             : //    }
    2375             : //}
    2376             : 
    2377             :     ///////////////////////
    2378             :     //                   //
    2379             :     // End Namespace avg //
    2380             :     //                   //
    2381             :     ///////////////////////
    2382             : 
    2383             : } // end avg
    2384             : 
    2385             : using namespace avg;
    2386             : 
    2387           0 : AveragingTvi2::AveragingTvi2 (ViImplementation2 * inputVi,
    2388           0 :                               const AveragingParameters & averagingParameters)
    2389             : : TransformingVi2 (inputVi),
    2390           0 :   averagingInterval_p (averagingParameters.getAveragingInterval()),
    2391           0 :   averagingOptions_p (averagingParameters.getOptions()),
    2392           0 :   averagingParameters_p (averagingParameters),
    2393           0 :   ddidLastUsed_p (-1),
    2394           0 :   inputViiAdvanced_p (false),
    2395           0 :   vbAvg_p (new VbAvg (averagingParameters, this))
    2396             : {
    2397           0 :     validateInputVi (inputVi);
    2398             : 
    2399             :     // Position input Vi to the first subchunk
    2400             : 
    2401           0 :     getVii()->originChunks();
    2402           0 :     getVii()->origin ();
    2403             : 
    2404           0 :     setVisBuffer (createAttachedVisBuffer (VbNoOptions));
    2405           0 : }
    2406             : 
    2407           0 : AveragingTvi2::~AveragingTvi2 ()
    2408             : {
    2409           0 :     delete vbAvg_p;
    2410           0 : }
    2411             : 
    2412             : //void
    2413             : //AveragingTvi2::advanceInputVii ()
    2414             : //{
    2415             : //    Assert (false);
    2416             : //
    2417             : //    // Attempt to advance to the next subchunk
    2418             : //
    2419             : //    getVii()->next ();
    2420             : //
    2421             : //    if (! getVii()->more()){
    2422             : //
    2423             : //        // No more subchunks so advance to the next chunk
    2424             : //
    2425             : //        getVii()->nextChunk();
    2426             : //
    2427             : //        if (! getVii()->moreChunks()){
    2428             : //            return; // no more chunks
    2429             : //        }
    2430             : //
    2431             : //        // Position to the first subchunk
    2432             : //
    2433             : //        getVii()->origin();
    2434             : //    }
    2435             : //}
    2436             : 
    2437             : //Int
    2438             : //AveragingTvi2::determineDdidToUse() const
    2439             : //{
    2440             : //    if (ddidLastUsed_p >= 0 && vbSet_p->anyAveragesReady (ddidLastUsed_p)){
    2441             : //        return ddidLastUsed_p;
    2442             : //    }
    2443             : //
    2444             : //    return vbSet_p->getFirstReadyDdid();
    2445             : //}
    2446             : 
    2447             : Bool
    2448           0 : AveragingTvi2::more () const
    2449             : {
    2450           0 :     return more_p;
    2451             : }
    2452             : 
    2453             : Bool
    2454           0 : AveragingTvi2::moreChunks () const
    2455             : {
    2456           0 :     return getVii()->moreChunks();
    2457             : }
    2458             : 
    2459             : void
    2460           0 : AveragingTvi2::next ()
    2461             : {
    2462           0 :     subchunkExists_p = false;
    2463             : 
    2464           0 :     startBuffer_p = endBuffer_p + 1;
    2465           0 :     endBuffer_p = startBuffer_p - 1;
    2466             : 
    2467           0 :     if (getVii()->more()){
    2468           0 :         getVii()->next();
    2469           0 :         endBuffer_p++;
    2470             :     }
    2471             : 
    2472           0 :     produceSubchunk ();
    2473             : 
    2474           0 :     subchunk_p.incrementSubChunk();
    2475           0 : }
    2476             : 
    2477             : void
    2478           0 : AveragingTvi2::nextChunk ()
    2479             : {
    2480             :     // New chunk, so toss all of the accumulators
    2481             : 
    2482           0 :     vbAvg_p->startChunk (getVii());
    2483             : 
    2484             :     // Advance the input to the next chunk as well.
    2485             : 
    2486           0 :     getVii()->nextChunk ();
    2487             : 
    2488           0 :     subchunk_p.incrementChunk();
    2489             : 
    2490           0 :     more_p = false;
    2491           0 : }
    2492             : 
    2493             : void
    2494           0 : AveragingTvi2::origin ()
    2495             : {
    2496             :     // Position input VI to the start of the chunk
    2497             : 
    2498           0 :     subchunk_p.resetSubChunk();
    2499             : 
    2500           0 :     getVii()->origin();
    2501             : 
    2502           0 :     startBuffer_p = 0;
    2503           0 :     endBuffer_p = -1;
    2504             : 
    2505             :     // Get the first subchunk ready.
    2506             : 
    2507           0 :     produceSubchunk ();
    2508           0 : }
    2509             : 
    2510             : void
    2511           0 : AveragingTvi2::originChunks (Bool forceRewind)
    2512             : {
    2513             :     // Ensure that the underlying VI is in a state where some metadata
    2514             :     // can be retrieved
    2515           0 :     getVii()->originChunks(forceRewind);
    2516             : 
    2517             :     // Initialize the chunk
    2518           0 :     vbAvg_p->startChunk (getVii());
    2519             : 
    2520           0 :     more_p = false;
    2521             : 
    2522           0 :     subchunk_p.resetToOrigin();
    2523           0 : }
    2524             : 
    2525             : /**
    2526             :  * Configure a VisBuffer with given averagingOptions related to phase shifting
    2527             :  *
    2528             :  * @param vb a VisBuffer to set up in terms of phase shifting
    2529             :  * @param averagingOptions averaging options enabled
    2530             :  * @param avgPars AveragingParmeters object to set into the buffer
    2531             :  */
    2532             : void
    2533           0 : setPhaseShiftingOptions(VisBuffer2 * vb, const AveragingOptions &averagingOptions,
    2534             :                         const AveragingParameters &avgPars)
    2535             : {
    2536           0 :     if (averagingOptions.contains (AveragingOptions::phaseShifting))
    2537             :     {
    2538           0 :         if (averagingOptions.contains (AveragingOptions::AverageObserved))
    2539             :         {
    2540           0 :             vb->visCube();
    2541             :         }
    2542             : 
    2543           0 :         if (averagingOptions.contains (AveragingOptions::AverageCorrected))
    2544             :         {
    2545           0 :             vb->visCubeCorrected();
    2546             :         }
    2547             : 
    2548           0 :         if (averagingOptions.contains (AveragingOptions::AverageModel))
    2549             :         {
    2550           0 :             vb->visCubeModel();
    2551             :         }
    2552             : 
    2553           0 :         vb->phaseCenterShift(avgPars.getXpcOffset(),
    2554             :                              avgPars.getYpcOffset());
    2555             :     }
    2556           0 : }
    2557             : 
    2558             : /**
    2559             :  * The iteration of this method is different depending on whether "row blocking" is used or
    2560             :  * not. The reason is that the while loop already had enough complexity embedded when fixes
    2561             :  * were done to get flagdata+time-average+row-blocking working (CAS-11910). Hopefully in the
    2562             :  * near future we can get rid of the hacky "row blocking" feature. For the time being, it is
    2563             :  * not clear how it could possibly work together with the "uvwdistance" feature. So better
    2564             :  * to keep those features separate.
    2565             :  *
    2566             :  * So the "if (block > 0)" separates iteration when using row blocking. That implies that
    2567             :  * row blocking takes precedence over (and disables) other features like
    2568             :  * "isUsingUvwDistance()".
    2569             :  * An alternative would be to add comparisons between block and vbToFill->appendSize() in
    2570             :  * the ifs below.  Something like:
    2571             :  *         if (! vbAvg_p->isUsingUvwDistance()
    2572             :  *            && (block == 0 && vbToFill->appendSize() > 0
    2573             :  *                || (block > 0 && vbToFill->appendSize() >= block)
    2574             :  *                )
    2575             :  *            ){
    2576             :  *          ...
    2577             :  *         else if ((block > 0 && vbToFill->appendSize() < block) ||
    2578             :  *                 vbToFill->appendSize() < nBaselines * nWindows){
    2579             :  *         ...
    2580             :  *         } else {
    2581             :  *
    2582             :  * But I prefer not adding more complexity to those ifs. The potential combinations would
    2583             :  * be too many to handle in a handful of if branches, and they are not well understood let
    2584             :  * alone well tested.
    2585             :  */
    2586             : void
    2587           0 : AveragingTvi2::produceSubchunk ()
    2588             : {
    2589           0 :     VisBufferImpl2 * vbToFill = dynamic_cast<VisBufferImpl2 *> (getVisBuffer());
    2590           0 :     assert (vbToFill != 0);
    2591             : 
    2592           0 :     vbToFill->setFillable (true); // New subchunk, so it's fillable
    2593             : 
    2594           0 :     vbAvg_p->setBufferToFill (vbToFill);
    2595             : 
    2596           0 :     Int nBaselines = nAntennas() * (nAntennas() -1) / 2;
    2597             :         // This is just a heuristic to keep output VBs from being too small
    2598             : 
    2599             :     // jagonzal: Handle nBaselines for SD case
    2600           0 :     if (nBaselines == 0) nBaselines = 1;
    2601             : 
    2602           0 :     auto block = getVii()->getRowBlocking();
    2603           0 :     while (getVii()->more()){
    2604             : 
    2605           0 :         VisBuffer2 * vb = getVii()->getVisBuffer();
    2606             : 
    2607           0 :         setPhaseShiftingOptions(vb, averagingOptions_p, averagingParameters_p);
    2608             : 
    2609           0 :         bool rowBlocking = block > 0;
    2610           0 :         vbAvg_p->accumulate (vb, subchunk_p);
    2611             : 
    2612           0 :         if (rowBlocking) {
    2613           0 :             auto app = vbToFill->appendSize();
    2614           0 :             if (app <= block) {
    2615           0 :                 getVii()->next();
    2616           0 :                 endBuffer_p++;
    2617             :             } else {
    2618           0 :                 break;
    2619             :             }
    2620             :         } else {
    2621           0 :             Int nWindows = vbAvg_p->nSpectralWindowsInBuffer ();
    2622           0 :             if (! vbAvg_p->isUsingUvwDistance() && vbToFill->appendSize() > 0){
    2623             :                 // Doing straight average and some data has been produced so
    2624             :                 // output it to the user
    2625           0 :                 break;
    2626             :             }
    2627           0 :             else if (vbToFill->appendSize() < nBaselines * nWindows) {
    2628           0 :                 getVii()->next();
    2629           0 :                 endBuffer_p++;
    2630             :             }
    2631             :             else {
    2632           0 :                 break;
    2633             :             }
    2634             :         }
    2635             :     };
    2636             : 
    2637           0 :     if (! getVii()->more()){
    2638           0 :         vbAvg_p->finalizeAverages ();
    2639             :     }
    2640             : 
    2641           0 :     vbAvg_p->finalizeBufferFilling ();
    2642             : 
    2643           0 :     more_p = getVii()->more() || // more to read
    2644           0 :              vbToFill->nRows() > 0; // some to process
    2645           0 : }
    2646             : 
    2647             : // Bool
    2648             : // AveragingTvi2::reachedAveragingBoundary()
    2649             : // {
    2650             : //     // An average can be terminated for a variety of reasons:
    2651             : //     // o the time interval has elapsed
    2652             : //     // o the current MS is completed
    2653             : //     // o no more input data
    2654             : //     // o other (e.g, change of scan, etc.)
    2655             : 
    2656             : //     Bool reachedIt = false;
    2657             : //     VisBuffer2 * vb = getVii()->getVisBuffer();
    2658             : 
    2659             : //     if (! getVii()->more()  && ! getVii ()->moreChunks()){
    2660             : 
    2661             : //         reachedIt = true; // no more input data
    2662             : //     }
    2663             : //     else if (vb->isNewMs()){
    2664             : //         reachedIt = true; // new MS
    2665             : //     }
    2666             : 
    2667             : //     return reachedIt;
    2668             : // }
    2669             : 
    2670             : //Bool
    2671             : //AveragingTvi2::subchunksReady() const
    2672             : //{
    2673             : //    Bool ready = vbSet_p->anyAveragesReady();
    2674             : //
    2675             : //    return ready;
    2676             : //}
    2677             : 
    2678             : void
    2679           0 : AveragingTvi2::validateInputVi (ViImplementation2 *)
    2680             : {
    2681             :     // Validate that the input VI is compatible with this VI.
    2682             : 
    2683             :     // No implmented right now
    2684           0 : }
    2685             : 
    2686           0 : Float AveragingTvi2::weightToSigma (Float weight)
    2687             : {
    2688           0 :     return weight > FLT_MIN ? 1.0 / std::sqrt (weight) : -1.0; // bogosity indicator
    2689             : }
    2690             : 
    2691             : Vector<Float>
    2692           0 : AveragingTvi2::average (const Matrix<Float> &data, const Matrix<Bool> &flags)
    2693             : {
    2694           0 :     IPosition shape = data.shape();
    2695           0 :     Vector<Float> result(shape(0),0);
    2696           0 :     Vector<uInt> samples(shape(0),0);
    2697           0 :     uInt nCorrelations = shape (0);
    2698           0 :     uInt nChannels = shape (1);
    2699             : 
    2700           0 :     for (uInt correlation = 0; correlation < nCorrelations; correlation++)
    2701             :     {
    2702           0 :         int nSamples = 0;
    2703           0 :         float sum = 0;
    2704           0 :         bool accumulatorFlag = true;
    2705             : 
    2706           0 :         for (uInt channel=0; channel< nChannels; channel++)
    2707             :         {
    2708           0 :             Bool inputFlag = flags(correlation,channel);
    2709             :             // true/true or false/false
    2710           0 :             if (accumulatorFlag == inputFlag)
    2711             :             {
    2712           0 :                 nSamples ++;
    2713           0 :                 sum += data (correlation, channel);
    2714             :             }
    2715             :             // true/false: Reset accumulation when accumulator switches from flagged to unflagged
    2716           0 :             else if ( accumulatorFlag and ! inputFlag )
    2717             :             {
    2718           0 :                 accumulatorFlag = false;
    2719           0 :                 nSamples = 1;
    2720           0 :                 sum = data (correlation, channel);
    2721             :             }
    2722             :             // else ignore case where accumulator is valid and data is not
    2723             : 
    2724             :         }
    2725             : 
    2726           0 :         result (correlation) = sum / (nSamples != 0 ? nSamples : 1);
    2727             :     }
    2728             : 
    2729           0 :     return result;
    2730           0 : }
    2731             : 
    2732             : Matrix<Float>
    2733           0 : AveragingTvi2::average (const Cube<Float> &data, const Cube<Bool> &flags)
    2734             : {
    2735           0 :         IPosition shape = data.shape();
    2736           0 :         uInt nRows = shape(2);
    2737           0 :         uInt nCorrelations = shape (0);
    2738             : 
    2739           0 :         Matrix<Float> result(nCorrelations, nRows, 0);
    2740             : 
    2741           0 :         for (uInt row=0; row < nRows; row++)
    2742             :         {
    2743           0 :                 result.column(row) = AveragingTvi2::average (data.xyPlane(row), flags.xyPlane(row));
    2744             :         }
    2745             : 
    2746           0 :     return result;
    2747           0 : }
    2748             : 
    2749             : /**
    2750             :  * Strategy to support different ways of propagating flags from the 'transformed' cube to
    2751             :  * the original ('propagated') cube. Iterates through rows, channels, correlations.
    2752             :  *
    2753             :  * This is meant to be used from propagateChanAvgFlags with at least two alternative
    2754             :  * functors. One to propagate flags as required by flagdata (preserve pre-existing flags
    2755             :  * in the original data cube), and a second one to propagate flags as required by plotms.
    2756             :  * CAS-12737, CAS-9985, CAS-12205.
    2757             :  *
    2758             :  * @param flagrow per row FLAG_ROW value
    2759             :  * @param flagMapped propagated FLAG_ROW
    2760             :  * @param flagCubeMapped Cube of flags in which flags are to be propagated
    2761             :  * @param row2AvgRow map of input_buffer_row_index->output_buffer_row_index (this is pre-
    2762             :  *        calculated by the "VbAvg" when averaging rows and is needed here).
    2763             :  */
    2764             : template <typename Functor>
    2765           0 : void cubePropagateFlags(const Vector<Bool> &flagRow,
    2766             :                         Vector<Bool> &flagMapped,
    2767             :                         Cube<Bool> &flagCubeMapped,
    2768             :                         std::vector<size_t> row2AvgRow,
    2769             :                         Functor propagate)
    2770             : {
    2771           0 :     Int nRowsMapped = flagCubeMapped.shape()[2];
    2772           0 :     for(Int rowMapped=0; rowMapped < nRowsMapped; ++rowMapped) {
    2773           0 :         size_t index = row2AvgRow[rowMapped];
    2774           0 :         flagMapped(rowMapped) = flagRow(index);
    2775             : 
    2776           0 :         for (Int chan_i=0; chan_i < flagCubeMapped.shape()[1]; ++chan_i) {
    2777           0 :             for (Int corr_i=0; corr_i<flagCubeMapped.shape()[0]; ++corr_i) {
    2778           0 :                 propagate(rowMapped, chan_i, corr_i, index);
    2779             :             }
    2780             :         }
    2781             :     }
    2782           0 : }
    2783             : 
    2784             : // -----------------------------------------------------------------------
    2785             : //
    2786             : // -----------------------------------------------------------------------
    2787           0 : void AveragingTvi2::writeFlag (const Cube<Bool> & flag)
    2788             : {
    2789             :     // Calculate FLAG_ROW from flag
    2790           0 :     Vector<Bool> flagRow;
    2791           0 :     TransformingVi2::calculateFlagRowFromFlagCube(flag, flagRow);
    2792             : 
    2793             :     const auto flagdataFlagPropagation =
    2794           0 :         averagingOptions_p.contains(AveragingOptions::flagdataFlagPropagation);
    2795             : 
    2796             :     // Propagate transformed flags
    2797           0 :     getVii()->origin();
    2798           0 :     Int currentBuffer = 0;
    2799           0 :     while (getVii()->more())
    2800             :     {
    2801           0 :         if ((currentBuffer >= startBuffer_p) and (currentBuffer <= endBuffer_p))
    2802             :         {
    2803             :             // Allocated propagated flag vector/cube
    2804           0 :             uInt nOriginalRows = getVii()->getVisBuffer()->nRows();
    2805           0 :             Vector<Bool> flagMapped(nOriginalRows,false);
    2806           0 :             Cube<Bool> flagCubeMapped;
    2807           0 :             flagCubeMapped = getVii()->getVisBuffer()->flagCube();
    2808             : 
    2809             :             // Keeping two separate blocks for 'flagdataFlagPropagation' (CAS-12737,
    2810             :             // CAS-12205, CAS-9985) until this issue is better settled.
    2811             :             // Fill propagated flag vector/cube
    2812           0 :             if (flagdataFlagPropagation)
    2813             :             {
    2814           0 :                 cubePropagateFlags(flagRow, flagMapped, flagCubeMapped, vbAvg_p->row2AvgRow(),
    2815           0 :                                    [&flag, &flagCubeMapped]
    2816           0 :                                    (uInt rowMapped, uInt chan_i, uInt corr_i, uInt index) {
    2817           0 :                                        if (flag(corr_i,chan_i,index))
    2818           0 :                                            flagCubeMapped(corr_i,chan_i,rowMapped) = true;
    2819           0 :                                    });
    2820             :             }
    2821             :             else
    2822             :             {
    2823           0 :                 cubePropagateFlags(flagRow, flagMapped, flagCubeMapped, vbAvg_p->row2AvgRow(),
    2824           0 :                                    [&flag, &flagCubeMapped]
    2825           0 :                                    (uInt rowMapped, uInt chan_i, uInt corr_i, uInt index) {
    2826           0 :                                        flagCubeMapped(corr_i, chan_i, rowMapped) =
    2827           0 :                                            flag(corr_i, chan_i, index);
    2828           0 :                                    });
    2829             :             }
    2830             : 
    2831             :             // Write propagated flag vector/cube
    2832           0 :             getVii()->writeFlag(flagCubeMapped);
    2833           0 :             getVii()->writeFlagRow(flagMapped);
    2834           0 :         }
    2835             : 
    2836           0 :         currentBuffer++;
    2837           0 :         getVii()->next();
    2838           0 :         if (currentBuffer > endBuffer_p) break;
    2839             :     }
    2840           0 : }
    2841             : 
    2842             : // -----------------------------------------------------------------------
    2843             : //
    2844             : // -----------------------------------------------------------------------
    2845           0 : void AveragingTvi2::writeFlagRow (const Vector<Bool> & flagRow)
    2846             : {
    2847             :         // Create index map for averaged data
    2848           0 :         VisBuffer2 *avgVB = getVisBuffer();
    2849           0 :         Vector<Int> avgAnt1 = avgVB->antenna1();
    2850           0 :         Vector<Int> avgAnt2 = avgVB->antenna2();
    2851           0 :         Vector<Int> avgSPW = avgVB->spectralWindows();
    2852             : 
    2853           0 :         std::map< Int, std::map <Int, std::map< Int, uInt> >  > spwAnt1Ant2IndexMap;
    2854           0 :         for (uInt avgRow=0;avgRow<avgAnt1.size();avgRow++)
    2855             :         {
    2856           0 :                 spwAnt1Ant2IndexMap[avgSPW(avgRow)][avgAnt1(avgRow)][avgAnt2(avgRow)] = avgRow;
    2857             :         }
    2858             : 
    2859             :         // Propagate transformed flags
    2860           0 :         getVii()->origin();
    2861           0 :         Int currentBuffer = 0;
    2862           0 :         while (getVii()->more())
    2863             :         {
    2864           0 :                 if ((currentBuffer >= startBuffer_p) and (currentBuffer <= endBuffer_p))
    2865             :                 {
    2866             :                         // Allocated propagated flag vector/cube
    2867           0 :                         uInt nOriginalRows = getVii()->getVisBuffer()->nRows();
    2868           0 :                         Vector<Bool> flagMapped(nOriginalRows,false);
    2869             : 
    2870             :                         // Get original ant1/ant2/spw cols. to determine the mapped index
    2871           0 :                         Vector<Int> orgAnt1 = getVii()->getVisBuffer()->antenna1();
    2872           0 :                         Vector<Int> orgAnt2 = getVii()->getVisBuffer()->antenna2();
    2873           0 :                         Vector<Int> orgSPW = getVii()->getVisBuffer()->spectralWindows();
    2874             : 
    2875             :                         // Fill propagated flag vector/cube
    2876           0 :                         for (uInt row=0;row<nOriginalRows;row++)
    2877             :                         {
    2878           0 :                                 uInt index = spwAnt1Ant2IndexMap[orgSPW(row)][orgAnt1(row)][orgAnt2(row)];
    2879           0 :                                 flagMapped(row) = flagRow(index);
    2880             :                         }
    2881             : 
    2882             :                         // Write propagated flag vector/cube
    2883           0 :                         getVii()->writeFlagRow(flagMapped);
    2884             : 
    2885           0 :                 }
    2886             : 
    2887           0 :                 currentBuffer += 1;
    2888           0 :                 getVii()->next();
    2889           0 :                 if (currentBuffer > endBuffer_p) break;
    2890             :         }
    2891             : 
    2892           0 :         return;
    2893           0 : }
    2894             : 
    2895           0 : void AveragingTvi2::visibilityObserved(casacore::Cube<casacore::Complex>& vis) const
    2896             : {
    2897           0 :   if(!averagingOptions_p.contains(AveragingOptions::AverageObserved))
    2898           0 :       throw AipsError("Requesting visibilityCorrected but column has not been averaged");
    2899           0 :     VisBuffer2* vb = getVisBuffer();
    2900           0 :     vis = vb->visCube();
    2901           0 :     return;
    2902             : }
    2903             : 
    2904           0 : void AveragingTvi2::visibilityCorrected(casacore::Cube<casacore::Complex>& vis) const
    2905             : {
    2906           0 :     if(!getVii()->getVisBuffer()->existsColumn (VisBufferComponent2::VisibilityCubeCorrected) ||
    2907           0 :         !averagingOptions_p.contains(AveragingOptions::AverageCorrected))
    2908           0 :         throw AipsError("Requesting visibilityCorrected but column not "
    2909           0 :             "provided by underlying VI or column has not been averaged");
    2910           0 :     VisBuffer2* vb = getVisBuffer();
    2911           0 :     vis = vb->visCubeCorrected();
    2912           0 :     return;
    2913             : }
    2914             : 
    2915           0 : void AveragingTvi2::visibilityModel(casacore::Cube<casacore::Complex>& vis) const
    2916             : {
    2917           0 :     if(!getVii()->getVisBuffer()->existsColumn (VisBufferComponent2::VisibilityCubeModel) ||
    2918           0 :         !averagingOptions_p.contains(AveragingOptions::AverageModel))
    2919           0 :         throw AipsError("Requesting visibilityModel but column not "
    2920           0 :             "provided by underlying VI or column has not been averaged");
    2921           0 :     VisBuffer2* vb = getVisBuffer();
    2922           0 :     vis = vb->visCubeModel();
    2923           0 :     return;
    2924             : }
    2925             : 
    2926           0 : void AveragingTvi2::floatData(casacore::Cube<casacore::Float>& fcube) const
    2927             : {
    2928           0 :     if(!getVii()->getVisBuffer()->existsColumn (VisBufferComponent2::VisibilityCubeFloat) ||
    2929           0 :         !averagingOptions_p.contains(AveragingOptions::AverageFloat))
    2930           0 :         throw AipsError("Requesting floatData but column not "
    2931           0 :             "provided by underlying VI or column has not been averaged");
    2932           0 :     VisBuffer2* vb = getVisBuffer();
    2933           0 :     fcube = vb->visCubeFloat();
    2934           0 :     return;
    2935             : }
    2936             : 
    2937           0 : void AveragingTvi2::flag(casacore::Cube<casacore::Bool>& flags) const
    2938             : {
    2939           0 :     VisBuffer2* vb = getVisBuffer();
    2940           0 :     flags = vb->flagCube();
    2941           0 :     return;
    2942             : }
    2943             : 
    2944           0 : void AveragingTvi2::flagRow(casacore::Vector<casacore::Bool>& rowflags) const
    2945             : {
    2946           0 :     VisBuffer2* vb = getVisBuffer();
    2947           0 :     rowflags = vb->flagRow();
    2948           0 :     return;
    2949             : }
    2950             : 
    2951           0 : void AveragingTvi2::sigma(casacore::Matrix<casacore::Float> & sigmat) const
    2952             : {
    2953           0 :     VisBuffer2* vb = getVisBuffer();
    2954           0 :     sigmat = vb->sigma();
    2955           0 :     return;
    2956             : }
    2957             : 
    2958           0 : void AveragingTvi2::weight (casacore::Matrix<casacore::Float> & wtmat) const
    2959             : {
    2960           0 :     VisBuffer2* vb = getVisBuffer();
    2961           0 :     wtmat = vb->weight();
    2962           0 :     return;
    2963             : }
    2964             : 
    2965           0 : void AveragingTvi2::weightSpectrum (casacore::Cube<casacore::Float> & wtsp) const
    2966             : {
    2967           0 :     VisBuffer2* vb = getVisBuffer();
    2968           0 :     wtsp = vb->weightSpectrum();
    2969           0 :     return;
    2970             : }
    2971             : 
    2972           0 : void AveragingTvi2::sigmaSpectrum (casacore::Cube<casacore::Float> & sigsp) const
    2973             : {
    2974           0 :     VisBuffer2* vb = getVisBuffer();
    2975           0 :     sigsp = vb->sigmaSpectrum();
    2976           0 :     return;
    2977             : }
    2978             : 
    2979           0 : void AveragingTvi2::exposure (casacore::Vector<double> & expo) const
    2980             : {
    2981           0 :     VisBuffer2* vb = getVisBuffer();
    2982           0 :     expo = vb->exposure();
    2983           0 :     return;
    2984             : }
    2985             : 
    2986           0 : void AveragingTvi2::getRowIds (Vector<rownr_t> & rowids) const
    2987             : {
    2988           0 :     VisBuffer2* vb = getVisBuffer();
    2989           0 :     rowids = vb->rowIds();
    2990           0 :     return;
    2991             : }
    2992             : 
    2993           0 : void AveragingTvi2::time (casacore::Vector<double> & t) const
    2994             : {
    2995           0 :     VisBuffer2* vb = getVisBuffer();
    2996           0 :     t = vb->time();
    2997           0 :     return;
    2998             : }
    2999             : 
    3000           0 : void AveragingTvi2::timeInterval (casacore::Vector<double> & ti) const
    3001             : {
    3002           0 :     VisBuffer2* vb = getVisBuffer();
    3003           0 :     ti = vb->timeInterval();
    3004           0 :     return;
    3005             : }
    3006             : 
    3007           0 : void AveragingTvi2::timeCentroid (casacore::Vector<double> & t) const
    3008             : {
    3009           0 :     VisBuffer2* vb = getVisBuffer();
    3010           0 :     t = vb->timeCentroid();
    3011           0 :     return;
    3012             : }
    3013             : 
    3014           0 : void AveragingTvi2::uvw (casacore::Matrix<double> & uvwmat) const
    3015             : {
    3016           0 :     VisBuffer2* vb = getVisBuffer();
    3017           0 :     uvwmat = vb->uvw();
    3018           0 :     return;
    3019             : }
    3020             : 
    3021           0 : void AveragingTvi2::antenna1 (casacore::Vector<casacore::Int> & ant1) const
    3022             : {
    3023           0 :     VisBuffer2* vb = getVisBuffer();
    3024           0 :     ant1 = vb->antenna1();
    3025           0 :     return;
    3026             : }
    3027             : 
    3028           0 : void AveragingTvi2::antenna2 (casacore::Vector<casacore::Int> & ant2) const
    3029             : {
    3030           0 :     VisBuffer2* vb = getVisBuffer();
    3031           0 :     ant2 = vb->antenna2();
    3032           0 :     return;
    3033             : }
    3034             : 
    3035           0 : casacore::Bool AveragingTvi2::weightSpectrumExists () const
    3036             : {
    3037             :   //According to VbAvg::startChunk code comments,
    3038             :   //there is always an output weightSpectrum. See also CAS-11559.
    3039           0 :   return true;
    3040             : }
    3041             : 
    3042           0 : casacore::Bool AveragingTvi2::sigmaSpectrumExists () const
    3043             : {
    3044             :   //According to VbAvg::startChunk code comments,
    3045             :   //there is always an output sigmaSpectrum. See also CAS-11559.
    3046           0 :   return true;
    3047             : }
    3048             : 
    3049             : } // end namespace vi
    3050             : 
    3051             : using namespace casacore;
    3052             : } // end namespace casa

Generated by: LCOV version 1.16