Avoid auto for Eigen expressions. (#899)

While harmless in most cases, auto can delay evaluation
of expressions in unexpected ways.
So it is better to avoid auto for Eigen expressions.
https://eigen.tuxfamily.org/dox/TopicPitfalls.html

cartographer/internal/mapping_3d/local_trajectory_builder.cc

@@ -229,7 +229,7 @@ LocalTrajectoryBuilder::InsertIntoSubmap(
   }
   active_submaps_.InsertRangeData(filtered_range_data_in_local,
                                   gravity_alignment);
-  const auto rotational_scan_matcher_histogram =
+  const Eigen::VectorXf rotational_scan_matcher_histogram =
       scan_matching::RotationalScanMatcher::ComputeHistogram(
           sensor::TransformPointCloud(
               filtered_range_data_in_tracking.returns,

cartographer/io/outlier_removing_points_processor.cc

@@ -94,7 +94,7 @@ void OutlierRemovingPointsProcessor::ProcessInPhaseTwo(
     const Eigen::Vector3f delta = batch.points[i] - batch.origin;
     const float length = delta.norm();
     for (float x = 0; x < length; x += voxel_size_) {
-      const auto index =
+      const Eigen::Array3i index =
           voxels_.GetCellIndex(batch.origin + (x / length) * delta);
       if (voxels_.value(index).hits > 0) {
         ++voxels_.mutable_value(index)->rays;
@@ -108,7 +108,8 @@ void OutlierRemovingPointsProcessor::ProcessInPhaseThree(
   constexpr double kMissPerHitLimit = 3;
   std::unordered_set<int> to_remove;
   for (size_t i = 0; i < batch->points.size(); ++i) {
-    const auto voxel = voxels_.value(voxels_.GetCellIndex(batch->points[i]));
+    const VoxelData voxel =
+        voxels_.value(voxels_.GetCellIndex(batch->points[i]));
     if (!(voxel.rays < kMissPerHitLimit * voxel.hits)) {
       to_remove.insert(i);
     }

cartographer/io/probability_grid_points_processor.cc

@@ -134,9 +134,9 @@ std::unique_ptr<Image> DrawProbabilityGrid(
   }
   auto image = common::make_unique<Image>(cell_limits.num_x_cells,
                                           cell_limits.num_y_cells);
-  for (const auto& xy_index :
+  for (const Eigen::Array2i& xy_index :
        cartographer::mapping_2d::XYIndexRangeIterator(cell_limits)) {
-    const auto index = xy_index + *offset;
+    const Eigen::Array2i index = xy_index + *offset;
     constexpr uint8 kUnknownValue = 128;
     const uint8 value =
         probability_grid.IsKnown(index)

cartographer/io/xray_points_processor.cc

@@ -149,7 +149,8 @@ void XRayPointsProcessor::WriteVoxels(const Aggregation& aggregation,
   }
 
   // Returns the (x, y) pixel of the given 'index'.
-  const auto voxel_index_to_pixel = [this](const Eigen::Array3i& index) {
+  const auto voxel_index_to_pixel =
+      [this](const Eigen::Array3i& index) -> Eigen::Array2i {
     // We flip the y axis, since matrices rows are counted from the top.
     return Eigen::Array2i(bounding_box_.max()[1] - index[1],
                           bounding_box_.max()[2] - index[2]);

cartographer/mapping_3d/scan_matching/ceres_scan_matcher_test.cc

@@ -37,7 +37,7 @@ class CeresScanMatcherTest : public ::testing::Test {
       : hybrid_grid_(1.f),
         expected_pose_(
             transform::Rigid3d::Translation(Eigen::Vector3d(-1., 0., 0.))) {
-    for (const auto& point :
+    for (const Eigen::Vector3f& point :
          {Eigen::Vector3f(-3.f, 2.f, 0.f), Eigen::Vector3f(-4.f, 2.f, 0.f),
           Eigen::Vector3f(-5.f, 2.f, 0.f), Eigen::Vector3f(-6.f, 2.f, 0.f),
           Eigen::Vector3f(-6.f, 3.f, 1.f), Eigen::Vector3f(-6.f, 4.f, 2.f),

cartographer/mapping_3d/scan_matching/interpolated_grid_test.cc

@@ -29,7 +29,7 @@ class InterpolatedGridTest : public ::testing::Test {
  protected:
   InterpolatedGridTest()
       : hybrid_grid_(0.1f), interpolated_grid_(hybrid_grid_) {
-    for (const auto& point :
+    for (const Eigen::Vector3f& point :
          {Eigen::Vector3f(-3.f, 2.f, 0.f), Eigen::Vector3f(-4.f, 2.f, 0.f),
           Eigen::Vector3f(-5.f, 2.f, 0.f), Eigen::Vector3f(-6.f, 2.f, 0.f),
           Eigen::Vector3f(-6.f, 3.f, 1.f), Eigen::Vector3f(-6.f, 4.f, 2.f),

cartographer/mapping_3d/scan_matching/low_resolution_matcher.cc

@@ -24,7 +24,8 @@ std::function<float(const transform::Rigid3f&)> CreateLowResolutionMatcher(
     const HybridGrid* low_resolution_grid, const sensor::PointCloud* points) {
   return [=](const transform::Rigid3f& pose) {
     float score = 0.f;
-    for (const auto& point : sensor::TransformPointCloud(*points, pose)) {
+    for (const Eigen::Vector3f& point :
+         sensor::TransformPointCloud(*points, pose)) {
       // TODO(zhengj, whess): Interpolate the Grid to get better score.
       score += low_resolution_grid->GetProbability(
           low_resolution_grid->GetCellIndex(point));

cartographer/mapping_3d/scan_matching/real_time_correlative_scan_matcher_test.cc

@@ -39,7 +39,7 @@ class RealTimeCorrelativeScanMatcherTest : public ::testing::Test {
       : hybrid_grid_(0.1f),
         expected_pose_(Eigen::Vector3d(-1., 0., 0.),
                        Eigen::Quaterniond::Identity()) {
-    for (const auto& point :
+    for (const Eigen::Vector3f& point :
          {Eigen::Vector3f(-3.f, 2.f, 0.f), Eigen::Vector3f(-4.f, 2.f, 0.f),
           Eigen::Vector3f(-5.f, 2.f, 0.f), Eigen::Vector3f(-6.f, 2.f, 0.f),
           Eigen::Vector3f(-6.f, 3.f, 1.f), Eigen::Vector3f(-6.f, 4.f, 2.f),

cartographer/sensor/compressed_point_cloud_test.cc

@@ -106,7 +106,7 @@ TEST(CompressPointCloudTest, CompressesNoGaps) {
   EXPECT_EQ(decompressed.size(), recompressed.size());
 
   std::vector<float> x_coord;
-  for (const auto& p : compressed) {
+  for (const Eigen::Vector3f& p : compressed) {
     x_coord.push_back(p[0]);
   }
   std::sort(x_coord.begin(), x_coord.end());

cartographer/sensor/point_cloud.cc

@@ -48,7 +48,7 @@ TimedPointCloud TransformTimedPointCloud(const TimedPointCloud& point_cloud,
 PointCloud CropPointCloud(const PointCloud& point_cloud, const float min_z,
                           const float max_z) {
   PointCloud cropped_point_cloud;
-  for (const auto& point : point_cloud) {
+  for (const Eigen::Vector3f& point : point_cloud) {
     if (min_z <= point.z() && point.z() <= max_z) {
       cropped_point_cloud.push_back(point);
     }
@@ -59,7 +59,7 @@ PointCloud CropPointCloud(const PointCloud& point_cloud, const float min_z,
 TimedPointCloud CropTimedPointCloud(const TimedPointCloud& point_cloud,
                                     const float min_z, const float max_z) {
   TimedPointCloud cropped_point_cloud;
-  for (const auto& point : point_cloud) {
+  for (const Eigen::Vector4f& point : point_cloud) {
     if (min_z <= point.z() && point.z() <= max_z) {
       cropped_point_cloud.push_back(point);
     }

cartographer/sensor/range_data.cc

@@ -58,11 +58,11 @@ proto::RangeData ToProto(const RangeData& range_data) {
   proto::RangeData proto;
   *proto.mutable_origin() = transform::ToProto(range_data.origin);
   proto.mutable_returns()->Reserve(range_data.returns.size());
-  for (const auto& point : range_data.returns) {
+  for (const Eigen::Vector3f& point : range_data.returns) {
     *proto.add_returns() = transform::ToProto(point);
   }
   proto.mutable_misses()->Reserve(range_data.misses.size());
-  for (const auto& point : range_data.misses) {
+  for (const Eigen::Vector3f& point : range_data.misses) {
     *proto.add_misses() = transform::ToProto(point);
   }
   return proto;

cartographer/sensor/timed_point_cloud_data.cc

@@ -28,7 +28,7 @@ proto::TimedPointCloudData ToProto(
   proto.set_timestamp(common::ToUniversal(timed_point_cloud_data.time));
   *proto.mutable_origin() = transform::ToProto(timed_point_cloud_data.origin);
   proto.mutable_point_data()->Reserve(timed_point_cloud_data.ranges.size());
-  for (const auto& range : timed_point_cloud_data.ranges) {
+  for (const Eigen::Vector4f& range : timed_point_cloud_data.ranges) {
     *proto.add_point_data() = transform::ToProto(range);
   }
   return proto;

cartographer/sensor/voxel_filter.cc

@@ -81,7 +81,7 @@ PointCloudType FilterPointCloudUsingVoxels(
     const PointCloudType& point_cloud,
     mapping_3d::HybridGridBase<uint8>* voxels) {
   PointCloudType results;
-  for (const auto& point : point_cloud) {
+  for (const typename PointCloudType::value_type& point : point_cloud) {
     auto* const value =
         voxels->mutable_value(voxels->GetCellIndex(point.template head<3>()));
     if (*value == 0) {