// 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 <utility>

#include "MantidDataHandling/DefaultEventLoader.h"
#include "MantidDataHandling/LoadEventNexus.h"
#include "MantidDataHandling/ProcessBankData.h"

using namespace Mantid::DataObjects;

namespace Mantid::DataHandling {

ProcessBankData::ProcessBankData(DefaultEventLoader &m_loader, std::string entry_name, API::Progress *prog,
                                 std::shared_ptr<std::vector<uint32_t>> event_id,
                                 std::shared_ptr<std::vector<float>> event_time_of_flight, size_t numEvents,
                                 size_t startAt, std::shared_ptr<std::vector<uint64_t>> event_index,
                                 std::shared_ptr<BankPulseTimes> thisBankPulseTimes, bool have_weight,
                                 std::shared_ptr<std::vector<float>> event_weight, detid_t min_event_id,
                                 detid_t max_event_id)
    : Task(), m_loader(m_loader), entry_name(std::move(entry_name)),
      pixelID_to_wi_vector(m_loader.pixelID_to_wi_vector), pixelID_to_wi_offset(m_loader.pixelID_to_wi_offset),
      prog(prog), event_id(std::move(event_id)), event_time_of_flight(std::move(event_time_of_flight)),
      numEvents(numEvents), startAt(startAt), event_index(std::move(event_index)),
      thisBankPulseTimes(std::move(thisBankPulseTimes)), have_weight(have_weight),
      event_weight(std::move(event_weight)), m_min_id(min_event_id), m_max_id(max_event_id) {
  // Cost is approximately proportional to the number of events to process.
  m_cost = static_cast<double>(numEvents);
}

namespace {
// this assumes that last_pulse_index is already to the point of including this
// one so we only need to search forward
inline size_t getPulseIndex(const size_t event_index, const size_t last_pulse_index,
                            const std::shared_ptr<std::vector<uint64_t>> &event_index_vec) {
  if (last_pulse_index + 1 >= event_index_vec->size())
    return last_pulse_index;

  // linear search is used because it is more likely that the next pulse index
  // is the correct one to use the + last_pulse_index + 1 is because we are
  // confirm that the next index is bigger, not the current
  const auto event_index_end = event_index_vec->cend();
  auto event_index_iter = event_index_vec->cbegin() + last_pulse_index;

  while ((event_index < *event_index_iter) || (event_index >= *(event_index_iter + 1))) {
    event_index_iter++;

    // make sure not to go past the end
    if (event_index_iter + 1 == event_index_end)
      break;
  }
  return std::distance(event_index_vec->cbegin(), event_index_iter);
}
} // namespace

/** Run the data processing
 * FIXME/TODO - split run() into readable methods
 */
void ProcessBankData::run() { // override {
  // Local tof limits
  double my_shortest_tof = static_cast<double>(std::numeric_limits<uint32_t>::max()) * 0.1;
  double my_longest_tof = 0.;
  // A count of "bad" TOFs that were too high
  size_t badTofs = 0;
  size_t my_discarded_events(0);

  prog->report(entry_name + ": precount");
  // ---- Pre-counting events per pixel ID ----
  auto &outputWS = m_loader.m_ws;
  auto *alg = m_loader.alg;
  if (m_loader.precount) {

    std::vector<size_t> counts(m_max_id - m_min_id + 1, 0);
    for (size_t i = 0; i < numEvents; i++) {
      const auto thisId = detid_t((*event_id)[i]);
      if (thisId >= m_min_id && thisId <= m_max_id)
        counts[thisId - m_min_id]++;
    }

    // Now we pre-allocate (reserve) the vectors of events in each pixel
    // counted
    const size_t numEventLists = outputWS.getNumberHistograms();
    for (detid_t pixID = m_min_id; pixID <= m_max_id; ++pixID) {
      if (counts[pixID - m_min_id] > 0) {
        size_t wi = getWorkspaceIndexFromPixelID(pixID);
        // Find the the workspace index corresponding to that pixel ID
        // Allocate it
        if (wi < numEventLists) {
          outputWS.reserveEventListAt(wi, counts[pixID - m_min_id]);
        }
        if (alg->getCancel())
          break; // User cancellation
      }
    }
  }

  // Check for canceled algorithm
  if (alg->getCancel()) {
    return;
  }

  // Default pulse time (if none are found)
  const bool pulsetimesincreasing =
      std::is_sorted(thisBankPulseTimes->pulseTimes.cbegin(), thisBankPulseTimes->pulseTimes.cend());
  if (!std::is_sorted(event_index->cbegin(), event_index->cend()))
    throw std::runtime_error("Event index is not sorted");

  // And there are this many pulses
  const auto NUM_PULSES = thisBankPulseTimes->pulseTimes.size();
  prog->report(entry_name + ": filling events");

  // Will we need to compress?
  const bool compress = (alg->compressTolerance >= 0);

  // Which detector IDs were touched?
  std::vector<bool> usedDetIds;
  usedDetIds.assign(m_max_id - m_min_id + 1, false);

  const double TOF_MIN = alg->filter_tof_min;
  const double TOF_MAX = alg->filter_tof_max;

  for (std::size_t pulseIndex = getPulseIndex(startAt, 0, event_index); pulseIndex < NUM_PULSES; pulseIndex++) {
    // Save the pulse time at this index for creating those events
    const auto pulsetime = thisBankPulseTimes->pulseTimes[pulseIndex];
    const int logPeriodNumber = thisBankPulseTimes->periodNumbers[pulseIndex];
    const int periodIndex = logPeriodNumber - 1;

    const auto firstEventIndex = getFirstEventIndex(pulseIndex);
    if (firstEventIndex > numEvents)
      break;

    const auto lastEventIndex = getLastEventIndex(pulseIndex, NUM_PULSES);
    if (firstEventIndex == lastEventIndex)
      continue;
    else if (firstEventIndex > lastEventIndex) {
      std::stringstream msg;
      msg << "Something went really wrong: " << firstEventIndex << " > " << lastEventIndex << "| " << entry_name
          << " startAt=" << startAt << " numEvents=" << event_index->size() << " RAWINDICES=["
          << firstEventIndex + startAt << ",?)"
          << " pulseIndex=" << pulseIndex << " of " << event_index->size();
      throw std::runtime_error(msg.str());
    }

    for (std::size_t eventIndex = firstEventIndex; eventIndex < lastEventIndex; ++eventIndex) {
      // We cached a pointer to the vector<tofEvent> -> so retrieve it and add
      // the event
      const detid_t detId = (*event_id)[eventIndex];
      if (detId >= m_min_id && detId <= m_max_id) {
        // Create the tofevent
        const auto tof = static_cast<double>((*event_time_of_flight)[eventIndex]);
        // this is fancy for check if value is in range
        if ((tof - TOF_MIN) * (tof - TOF_MAX) <= 0.) {
          // Handle simulated data if present
          if (have_weight) {
            auto *eventVector = m_loader.weightedEventVectors[periodIndex][detId];
            // NULL eventVector indicates a bad spectrum lookup
            if (eventVector) {
              const auto weight = static_cast<double>((*event_weight)[eventIndex]);
              const double errorSq = weight * weight;
              eventVector->emplace_back(tof, pulsetime, weight, errorSq);
            } else {
              ++my_discarded_events;
            }
          } else {
            // We have cached the vector of events for this detector ID
            auto *eventVector = m_loader.eventVectors[periodIndex][detId];
            // NULL eventVector indicates a bad spectrum lookup
            if (eventVector) {
              eventVector->emplace_back(tof, pulsetime);
            } else {
              ++my_discarded_events;
            }
          }

          // Skip any events that are the cause of bad DAS data (e.g. a negative
          // number in uint32 -> 2.4 billion * 100 nanosec = 2.4e8 microsec)
          if (tof < 2e8) {
            // tof limits from things observed here
            if (tof > my_longest_tof) {
              my_longest_tof = tof;
            }
            if (tof < my_shortest_tof) {
              my_shortest_tof = tof;
            }
          } else
            badTofs++;

          // Track all the touched wi
          usedDetIds[detId - m_min_id] = true;
        } // valid time-of-flight

      } // valid detector IDs
    }   // for events in pulse
    // check if cancelled after each pulse
    if (alg->getCancel())
      break;
  } // for pulses

  // Check for canceled algorithm
  if (alg->getCancel()) {
    return;
  }

  //------------ Compress Events (or set sort order) ------------------
  // Do it on all the detector IDs we touched
  const size_t numEventLists = outputWS.getNumberHistograms();
  for (detid_t pixID = m_min_id; pixID <= m_max_id; ++pixID) {
    if (usedDetIds[pixID - m_min_id]) {
      // Find the the workspace index corresponding to that pixel ID
      size_t wi = getWorkspaceIndexFromPixelID(pixID);
      if (wi < numEventLists) {
        auto &el = outputWS.getSpectrum(wi);
        if (compress)
          el.compressEvents(alg->compressTolerance, &el);
        else {
          if (pulsetimesincreasing)
            el.setSortOrder(DataObjects::PULSETIME_SORT);
          else
            el.setSortOrder(DataObjects::UNSORTED);
        }
      }
    }
  }
  prog->report(entry_name + ": filled events");

  alg->getLogger().debug() << entry_name << (pulsetimesincreasing ? " had " : " DID NOT have ")
                           << "monotonically increasing pulse times\n";

  // Join back up the tof limits to the global ones
  // This is not thread safe, so only one thread at a time runs this.
  {
    std::lock_guard<std::mutex> _lock(alg->m_tofMutex);
    if (my_shortest_tof < alg->shortest_tof) {
      alg->shortest_tof = my_shortest_tof;
    }
    if (my_longest_tof > alg->longest_tof) {
      alg->longest_tof = my_longest_tof;
    }
    alg->bad_tofs += badTofs;
    alg->discarded_events += my_discarded_events;
  }

#ifndef _WIN32
  alg->getLogger().debug() << "Time to process " << entry_name << " " << m_timer << "\n";
#endif
} // END-OF-RUN()

size_t ProcessBankData::getFirstEventIndex(const size_t pulseIndex) const {
  const auto firstEventIndex = event_index->operator[](pulseIndex);
  if (firstEventIndex >= startAt)
    return firstEventIndex - startAt;
  else
    return 0;
}

size_t ProcessBankData::getLastEventIndex(const size_t pulseIndex, const size_t numPulses) const {
  if (pulseIndex + 1 >= numPulses)
    return numEvents;

  const size_t lastEventIndex = event_index->operator[](pulseIndex + 1) - startAt;

  return std::min(lastEventIndex, numEvents);
}

/**
 * Get the workspace index for a given pixel ID. Throws if the pixel ID is
 * not in the expected range.
 *
 * @param pixID :: The pixel ID to look up
 * @return The workspace index for this pixel
 */
size_t ProcessBankData::getWorkspaceIndexFromPixelID(const detid_t pixID) {
  // Check that the vector index is not out of range
  const detid_t offset_pixID = pixID + pixelID_to_wi_offset;
  if (offset_pixID < 0 || offset_pixID >= static_cast<int32_t>(pixelID_to_wi_vector.size())) {
    std::stringstream msg;
    msg << "Error finding workspace index; pixelID " << pixID << " with offset " << pixelID_to_wi_offset
        << " is out of range (length=" << pixelID_to_wi_vector.size() << ")";
    throw std::runtime_error(msg.str());
  }
  return pixelID_to_wi_vector[offset_pixID];
}
} // namespace Mantid::DataHandling