SparseWorkspace.cpp

Go to the documentation of this file.

// Mantid Repository : https://github.com/mantidproject/mantid
//
// Copyright © 2018 ISIS Rutherford Appleton Laboratory UKRI,
//   NScD Oak Ridge National Laboratory, European Spallation Source,
//   Institut Laue - Langevin & CSNS, Institute of High Energy Physics, CAS
// SPDX - License - Identifier: GPL - 3.0 +
#include "MantidAlgorithms/SampleCorrections/SparseWorkspace.h"
 
#include "MantidAPI/MatrixWorkspace.h"
#include "MantidAPI/Run.h"
#include "MantidAPI/SpectrumInfo.h"
#include "MantidDataObjects/WorkspaceCreation.h"
#include "MantidGeometry/Instrument.h"
#include "MantidGeometry/Instrument/ComponentInfo.h"
#include "MantidGeometry/Instrument/Detector.h"
#include "MantidGeometry/Instrument/DetectorInfo.h"
#include "MantidGeometry/Instrument/ReferenceFrame.h"
#include "MantidGeometry/Objects/CSGObject.h"
#include "MantidGeometry/Objects/ShapeFactory.h"
#include "MantidHistogramData/HistogramIterator.h"
#include "MantidKernel/Logger.h"
 
#include <Poco/DOM/AutoPtr.h>
#include <Poco/DOM/Document.h>
 
namespace {
bool constantIndirectEFixed(const Mantid::API::ExperimentInfo &info, const std::vector<Mantid::detid_t> &detIDs) {
  if (detIDs.empty())
    return true;
  const auto e = info.getEFixed(detIDs[0]);
  return std::all_of(detIDs.cbegin() + 1, detIDs.cend(), [&](const auto &detID) { return e == info.getEFixed(detID); });
}
 
constexpr double R = 1.0; // This will be the default L2 distance.
 
Mantid::Kernel::Logger g_log("SparseWorkspace");
} // namespace
 
namespace Mantid::Algorithms {
 
SparseWorkspace::SparseWorkspace(const API::MatrixWorkspace &modelWS, const size_t wavelengthPoints, const size_t rows,
                                 const size_t columns)
    : Workspace2D() {
  double minLat, maxLat, minLong, maxLong;
  std::tie(minLat, maxLat, minLong, maxLong) = extremeAngles(modelWS);
  m_gridDef = std::make_unique<Algorithms::DetectorGridDefinition>(minLat, maxLat, rows, minLong, maxLong, columns);
  if ((rows < 3) || (columns < 3)) {
    g_log.warning("Can't calculate errors on a sparse workspace with lat or "
                  "long dimension < 3");
  }
  const size_t numSpectra = rows * columns;
  const auto h = modelHistogram(modelWS, wavelengthPoints);
  initialize(numSpectra, h);
 
  // Build a quite standard and somewhat complete instrument.
  auto instrument = std::make_shared<Geometry::Instrument>("MC_simulation_instrument");
  const auto refFrame = modelWS.getInstrument()->getReferenceFrame();
 
  instrument->setReferenceFrame(std::make_shared<Geometry::ReferenceFrame>(*refFrame));
  // The sparse instrument is build around origin.
  constexpr Kernel::V3D samplePos{0.0, 0.0, 0.0};
  auto sample = std::make_unique<Geometry::Component>("sample", instrument.get());
  sample->setPos(samplePos);
  instrument->add(sample.get());
  instrument->markAsSamplePos(sample.release());
  // Add source behind the sample.
  const Kernel::V3D sourcePos = [&]() {
    Kernel::V3D p;
    p[refFrame->pointingAlongBeam()] = -2.0 * R;
    return p;
  }();
  auto source = std::make_unique<Geometry::ObjComponent>("source", nullptr, instrument.get());
  source->setPos(sourcePos);
  instrument->add(source.get());
  instrument->markAsSource(source.release());
 
  auto detShape = makeCubeShape();
  for (size_t col = 0; col < columns; ++col) {
    const auto lon = m_gridDef->longitudeAt(col);
    for (size_t row = 0; row < rows; ++row) {
      const auto lat = m_gridDef->latitudeAt(row);
      const size_t index = col * rows + row;
      const auto detID = static_cast<int>(index);
      std::ostringstream detName;
      detName << "det-" << detID;
      auto det = std::make_unique<Geometry::Detector>(detName.str(), detID, detShape, instrument.get());
      const Kernel::V3D pos = [&]() {
        Kernel::V3D p;
        p[refFrame->pointingHorizontal()] = R * std::sin(lon) * std::cos(lat);
        p[refFrame->pointingUp()] = R * std::sin(lat);
        p[refFrame->pointingAlongBeam()] = R * std::cos(lon) * std::cos(lat);
        return p;
      }();
      det->setPos(pos);
      getSpectrum(index).setDetectorID(detID);
      instrument->add(det.get());
      instrument->markAsDetector(det.release());
    }
  }
  setInstrument(instrument);
 
  // Copy things needed for the simulation from the model workspace.
  auto &componentInfo = mutableComponentInfo();
  const auto &modelComponentInfo = modelWS.componentInfo();
  // Copy beam parameters.
  const auto modelSource = modelComponentInfo.source();
  const auto sourceIndex = componentInfo.source();
  const auto beamShapeParam = modelComponentInfo.getParameterAsString(modelSource, "beam-shape");
  if (!beamShapeParam.empty())
    componentInfo.addString(sourceIndex, "beam-shape", beamShapeParam);
  const auto beamWidthParam = modelComponentInfo.getNumberParameter(modelSource, "beam-width");
  const auto beamHeightParam = modelComponentInfo.getNumberParameter(modelSource, "beam-height");
  if (beamWidthParam.size() == 1 && beamHeightParam.size() == 1) {
    componentInfo.addDouble(sourceIndex, "beam-width", beamWidthParam[0]);
    componentInfo.addDouble(sourceIndex, "beam-height", beamHeightParam[0]);
  }
  const auto beamRadiusParam = modelComponentInfo.getNumberParameter(modelSource, "beam-radius");
  if (beamRadiusParam.size() == 1) {
    componentInfo.addDouble(sourceIndex, "beam-radius", beamRadiusParam[0]);
  }
  // Add information about EFixed in a proper place.
  const auto eMode = modelWS.getEMode();
  mutableRun().addProperty("deltaE-mode", Kernel::DeltaEMode::asString(eMode));
  if (eMode == Kernel::DeltaEMode::Direct) {
    mutableRun().addProperty("Ei", modelWS.getEFixed());
  } else if (eMode == Kernel::DeltaEMode::Indirect) {
    std::vector<detid_t> detIDs;
    auto modelInstrument = modelWS.getInstrument();
    for (size_t i = 0; i < modelWS.getNumberHistograms(); i++) {
      const auto &spec = modelWS.getSpectrum(i);
      const auto &specDetIDs = spec.getDetectorIDs();
      if (!modelInstrument->isMonitor(specDetIDs))
        detIDs.insert(detIDs.end(), specDetIDs.begin(), specDetIDs.end());
    }
    if (detIDs.empty()) {
      throw std::runtime_error("Sparse instrument: no non-monitor detectors found for Indirect mode.");
    }
    if (!constantIndirectEFixed(modelWS, detIDs)) {
      throw std::runtime_error("Sparse instrument with variable EFixed not supported.");
    }
    const auto e = modelWS.getEFixed(detIDs[0]);
    const auto &sparseDetIDs = detectorInfo().detectorIDs();
    for (int sparseDetID : sparseDetIDs) {
      setEFixed(sparseDetID, e);
    }
  }
}
 
SparseWorkspace::SparseWorkspace(const SparseWorkspace &other)
    : Workspace2D(other), m_gridDef(std::make_unique<Algorithms::DetectorGridDefinition>(*other.m_gridDef)) {}
 
std::tuple<double, double, double, double> SparseWorkspace::extremeAngles(const API::MatrixWorkspace &ws) {
  const auto &spectrumInfo = ws.spectrumInfo();
  double minLat = std::numeric_limits<double>::max();
  double maxLat = std::numeric_limits<double>::lowest();
  double minLong = std::numeric_limits<double>::max();
  double maxLong = std::numeric_limits<double>::lowest();
  for (size_t i = 0; i < ws.getNumberHistograms(); ++i) {
    if (spectrumInfo.hasDetectors(i)) {
      double lat, lon;
      std::tie(lat, lon) = spectrumInfo.geographicalAngles(i);
      if (lat < minLat) {
        minLat = lat;
      }
      if (lat > maxLat) {
        maxLat = lat;
      }
      if (lon < minLong) {
        minLong = lon;
      }
      if (lon > maxLong) {
        maxLong = lon;
      }
    }
  }
  return std::make_tuple(minLat, maxLat, minLong, maxLong);
}
 
std::tuple<double, double> SparseWorkspace::extremeWavelengths(const API::MatrixWorkspace &ws) {
  double currentMin = std::numeric_limits<double>::max();
  double currentMax = std::numeric_limits<double>::lowest();
  for (size_t i = 0; i < ws.getNumberHistograms(); ++i) {
    const auto h = ws.histogram(i);
    const auto first = h.begin();
    currentMin = std::min(first->center(), currentMin);
    const auto last = std::prev(h.end());
    currentMax = std::max(last->center(), currentMax);
  }
  return std::make_tuple(currentMin, currentMax);
}
 
Mantid::HistogramData::Histogram SparseWorkspace::modelHistogram(const API::MatrixWorkspace &modelWS,
                                                                 const size_t wavelengthPoints) {
  double minWavelength, maxWavelength;
  std::tie(minWavelength, maxWavelength) = extremeWavelengths(modelWS);
  HistogramData::Frequencies ys(wavelengthPoints, 0.0);
  HistogramData::FrequencyVariances es(wavelengthPoints, 0.0);
  HistogramData::Points ps(wavelengthPoints, 0.0);
  HistogramData::Histogram h(ps, ys, es);
  auto &xs = h.mutableX();
  if (wavelengthPoints > 1) {
    const double step = (maxWavelength - minWavelength) / static_cast<double>(wavelengthPoints - 1);
    for (size_t i = 0; i < xs.size() - 1; ++i) {
      xs[i] = minWavelength + step * static_cast<double>(i);
    }
    // Force last point as otherwise it might be slightly off due to
    // small rounding errors in the calculation above.
    xs.back() = maxWavelength;
  } else {
    xs.front() = (minWavelength + maxWavelength) / 2.0;
  }
  return h;
}
 
Geometry::IObject_sptr SparseWorkspace::makeCubeShape() {
  using namespace Poco::XML;
  const double dimension = 0.05;
  AutoPtr<Document> shapeDescription = new Document;
  AutoPtr<Element> typeElement = shapeDescription->createElement("type");
  typeElement->setAttribute("name", "detector");
  AutoPtr<Element> shapeElement = shapeDescription->createElement("cuboid");
  shapeElement->setAttribute("id", "cube");
  const std::string posCoord = std::to_string(dimension / 2);
  const std::string negCoord = std::to_string(-dimension / 2);
  AutoPtr<Element> element = shapeDescription->createElement("left-front-bottom-point");
  element->setAttribute("x", negCoord);
  element->setAttribute("y", negCoord);
  element->setAttribute("z", posCoord);
  shapeElement->appendChild(element);
  element = shapeDescription->createElement("left-front-top-point");
  element->setAttribute("x", negCoord);
  element->setAttribute("y", posCoord);
  element->setAttribute("z", posCoord);
  shapeElement->appendChild(element);
  element = shapeDescription->createElement("left-back-bottom-point");
  element->setAttribute("x", negCoord);
  element->setAttribute("y", negCoord);
  element->setAttribute("z", negCoord);
  shapeElement->appendChild(element);
  element = shapeDescription->createElement("right-front-bottom-point");
  element->setAttribute("x", posCoord);
  element->setAttribute("y", negCoord);
  element->setAttribute("z", posCoord);
  shapeElement->appendChild(element);
  typeElement->appendChild(shapeElement);
  AutoPtr<Element> algebraElement = shapeDescription->createElement("algebra");
  algebraElement->setAttribute("val", "cube");
  typeElement->appendChild(algebraElement);
  Geometry::ShapeFactory shapeFactory;
  return shapeFactory.createShape(typeElement);
}
 
double SparseWorkspace::greatCircleDistance(const double lat1, const double long1, const double lat2,
                                            const double long2) {
  const double latD = std::sin((lat2 - lat1) / 2.0);
  const double longD = std::sin((long2 - long1) / 2.0);
  const double S = latD * latD + std::cos(lat1) * std::cos(lat2) * longD * longD;
  return 2.0 * std::asin(std::sqrt(S));
}
 
std::array<double, 4> SparseWorkspace::inverseDistanceWeights(const std::array<double, 4> &distances) {
  std::array<double, 4> weights;
  for (size_t i = 0; i < weights.size(); ++i) {
    if (distances[i] == 0.0) {
      weights.fill(0.0);
      weights[i] = 1.0;
      return weights;
    }
    weights[i] = 1.0 / distances[i] / distances[i];
  }
  return weights;
}
 
HistogramData::HistogramY SparseWorkspace::secondDerivative(const std::array<size_t, 3> &indices,
                                                            const double distanceStep) const {
  HistogramData::HistogramY avgSecondDeriv(blocksize());
 
  HistogramData::HistogramY sumSecondDeriv(blocksize());
  auto firstDerivB = (y(indices[2]) - y(indices[1])) / distanceStep;
  auto firstDerivA = (y(indices[1]) - y(indices[0])) / distanceStep;
  auto secondDerivA = (firstDerivB - firstDerivA) / distanceStep;
 
  return secondDerivA;
}
 
HistogramData::Histogram SparseWorkspace::interpolateFromDetectorGrid(const double lat, const double lon) const {
  const auto indices = m_gridDef->nearestNeighbourIndices(lat, lon);
 
  auto h = histogram(0);
 
  std::array<double, 4> distances;
  for (size_t i = 0; i < 4; ++i) {
    double detLat, detLong;
    std::tie(detLat, detLong) = spectrumInfo().geographicalAngles(indices[i]);
    distances[i] = greatCircleDistance(lat, lon, detLat, detLong);
  }
  const auto weights = inverseDistanceWeights(distances);
  auto weightSum = weights[0];
  h.mutableY() = weights[0] * y(indices[0]);
  for (size_t i = 1; i < 4; ++i) {
    weightSum += weights[i];
    h.mutableY() += weights[i] * y(indices[i]);
  }
  h.mutableY() /= weightSum;
  return h;
}
 
HistogramData::HistogramE SparseWorkspace::esq(const HistogramData::HistogramE &e) const { return e * e; }
 
HistogramData::HistogramE SparseWorkspace::esqrt(HistogramData::HistogramE e) const {
  std::transform(e.cbegin(), e.cend(), e.begin(), [](double f) -> double { return sqrt(f); });
  return e;
}
 
HistogramData::Histogram SparseWorkspace::bilinearInterpolateFromDetectorGrid(const double lat,
                                                                              const double lon) const {
 
  size_t nearestLatIndex, nearestLonIndex;
  std::tie(nearestLatIndex, nearestLonIndex) = m_gridDef->getNearestVertex(lat, lon);
 
  std::array<std::array<size_t, 2>, 2> detIndices;
  for (size_t i = 0; i < 2; i++) {
    for (size_t j = 0; j < 2; j++) {
      detIndices[i][j] = m_gridDef->getDetectorIndex(nearestLatIndex + j, nearestLonIndex + i);
    }
  }
 
  double latLow, longLow, latHigh, longHigh;
  std::tie(latLow, longLow) = spectrumInfo().geographicalAngles(detIndices[0][0]);
  std::tie(latHigh, longHigh) = spectrumInfo().geographicalAngles(detIndices[1][1]);
 
  // interpolate across different longitudes
  auto ylat1 = ((longHigh - lon) * y(detIndices[0][0]) + (lon - longLow) * y(detIndices[1][0])) / (longHigh - longLow);
  auto errLat1 = esqrt(esq((longHigh - lon) * e(detIndices[0][0])) + esq((lon - longLow) * e(detIndices[1][0]))) /
                 (longHigh - longLow);
 
  auto ylat2 = ((longHigh - lon) * y(detIndices[0][1]) + (lon - longLow) * y(detIndices[1][1])) / (longHigh - longLow);
  auto errLat2 = esqrt(esq((longHigh - lon) * e(detIndices[0][1])) + esq((lon - longLow) * e(detIndices[1][1]))) /
                 (longHigh - longLow);
 
  // interpolate across different latitudes
  auto interpY = ((latHigh - lat) * ylat1 + (lat - latLow) * ylat2) / (latHigh - latLow);
  auto errFromSourcePoints = esqrt(esq((latHigh - lat) * errLat1) + esq((lat - latLow) * errLat2)) / (latHigh - latLow);
 
  // calculate interpolation errors if possible
  HistogramData::HistogramE interpolationError(e(0).size(), 0);
  if ((m_gridDef->numberColumns() > 2) && (m_gridDef->numberRows() > 2)) {
    // step back a further row\col (if possible) to estimate the second
    // derivative
    auto nearestLatIndexSec = nearestLatIndex > 0 ? nearestLatIndex - 1 : 0;
    auto nearestLonIndexSec = nearestLonIndex > 0 ? nearestLonIndex - 1 : 0;
    std::array<size_t, 3> threeIndices;
 
    // 2nd derivative in longitude
    for (int i = 0; i < 3; i++) {
      threeIndices[i] = m_gridDef->getDetectorIndex(nearestLatIndex, nearestLonIndexSec + i);
    }
    auto avgSecondDerivLong = secondDerivative(threeIndices, longHigh - longLow);
 
    // 2nd derivative in latitude
    for (int i = 0; i < 3; i++) {
      threeIndices[i] = m_gridDef->getDetectorIndex(nearestLatIndexSec + i, nearestLonIndex);
    }
    auto avgSecondDerivLat = secondDerivative(threeIndices, latHigh - latLow);
 
    // calculate the interpolation error according to a Taylor expansion from
    // the low lat and low long points. Doesn't make any difference if do this
    // in angle or distance because scaling factor (~ radius) cancels out
    HistogramData::HistogramY interpolationErrorAsYData =
        0.5 * (lon - longLow) * (longHigh - lon) * avgSecondDerivLong +
        0.5 * (lat - latLow) * (latHigh - lat) * avgSecondDerivLat;
    // convert from HistogramY to HistogramE
    interpolationError = interpolationErrorAsYData.rawData();
  }
 
  auto h = histogram(0);
  h.mutableY() = interpY;
  // combine the (independent) errors
  // Note - interpolationError may be negative if the 2nd derivative is negative
  // so the following line also has useful side effect of taking abs value
  h.mutableE() = esqrt(esq(interpolationError) + esq(errFromSourcePoints));
  return h;
}
 
SparseWorkspace *SparseWorkspace::doClone() const { return new SparseWorkspace(*this); }
 
} // namespace Mantid::Algorithms