Refactor _readSelection.

src/population.hpp

@@ -20,6 +20,8 @@
 
 #include <fmt/format.h>
 
+#include "read_bulk.hpp"
+#include "read_canonical_selection.hpp"
 #include <highfive/H5File.hpp>
 
 namespace bbp {
@@ -58,45 +60,22 @@ std::set<std::string> _listExplicitEnumerations(const HighFive::Group h5Group,
 }
 
 template <typename T>
-std::vector<T> _readChunk(const HighFive::DataSet& dset, const Selection::Range& range) {
- std::vector<T> result;
- assert(range.first < range.second);
- auto chunkSize = static_cast<size_t>(range.second - range.first);
- dset.select({static_cast<size_t>(range.first)}, {chunkSize}).read(result);
- return result;
-}
-
-
-template <typename T, typename std::enable_if<!std::is_pod<T>::value>::type* = nullptr>
 std::vector<T> _readSelection(const HighFive::DataSet& dset, const Selection& selection) {
- if (selection.ranges().empty() || dset.getElementCount() == 0) {
+ if (dset.getElementCount() == 0) {
  return {};
  }
 
- if (selection.ranges().size() == 1) {
- return _readChunk<T>(dset, selection.ranges().front());
+ if (bulk_read::detail::isCanonical(selection)) {
+ return detail::readCanonicalSelection<T>(dset, selection);
  }
 
- std::vector<T> result;
-
- // for POD types we can pre-allocate result vector... see below template specialization
- for (const auto& range : selection.ranges()) {
- for (auto& x : _readChunk<T>(dset, range)) {
- result.emplace_back(std::move(x));
- }
- }
-
- return result;
-}
-
-
-template <typename T, typename std::enable_if<std::is_pod<T>::value>::type* = nullptr>
-std::vector<T> _readSelection(const HighFive::DataSet& dset, const Selection& selection) {
- if (selection.ranges().empty() || dset.getElementCount() == 0) {
- return {};
- } else if (selection.ranges().size() == 1) {
- return _readChunk<T>(dset, selection.ranges().front());
- }
+ // The fully general case:
+ //
+ // 1. Create a canonical selection and read into `linear_result`.
+ // 2. Copy values from the canonical `linear_results` to their final
+ // destination.
+ auto canonicalRanges = bulk_read::sortAndMerge(selection, 0);
+ auto linear_result = detail::readCanonicalSelection<T>(dset, canonicalRanges);
 
  const auto ids = selection.flatten();
 
@@ -106,19 +85,15 @@ std::vector<T> _readSelection(const HighFive::DataSet& dset, const Selection& se
  return ids[i0] < ids[i1];
  });
 
- std::vector<std::size_t> ids_sorted;
- ids_sorted.reserve(ids.size());
- std::transform(ids_index.begin(),
- ids_index.end(),
- std::back_inserter(ids_sorted),
- [&ids](size_t i) { return static_cast<size_t>(ids[i]); });
-
- std::vector<T> linear_result;
- dset.select(HighFive::ElementSet{ids_sorted}).read(linear_result);
+ std::vector<T> result(ids.size());
+ size_t linear_index = 0;
+ result[ids_index[0]] = linear_result[0];
+ for (size_t i = 1; i < ids.size(); ++i) {
+ if (ids[ids_index[i - 1]] != ids[ids_index[i]]) {
+ linear_index += 1;
+ }
 
- std::vector<T> result(ids_sorted.size());
- for (size_t i = 0; i < ids_sorted.size(); ++i) {
- result[ids_index[i]] = linear_result[i];
+ result[ids_index[i]] = linear_result[linear_index];
  }
 
  return result;

src/read_canonical_selection.hpp

@@ -0,0 +1,35 @@
+#pragma once
+
+#include <vector>
+
+#include <highfive/H5File.hpp>
+
+namespace bbp {
+namespace sonata {
+namespace detail {
+
+
+template <class Range>
+HighFive::HyperSlab make_hyperslab(const std::vector<Range>& ranges) {
+ HighFive::HyperSlab slab;
+ for (const auto& range : ranges) {
+ size_t i_begin = std::get<0>(range);
+ size_t i_end = std::get<1>(range);
+ slab |= HighFive::RegularHyperSlab({i_begin}, {i_end - i_begin});
+ }
+
+ return slab;
+}
+
+template <class T>
+std::vector<T> readCanonicalSelection(const HighFive::DataSet& dset, const Selection& selection) {
+ if (selection.empty()) {
+ return {};
+ }
+
+ return dset.select(make_hyperslab(selection.ranges())).template read<std::vector<T>>();
+}
+
+} // namespace detail
+} // namespace sonata
+} // namespace bbp