Add PushCSC for SparsePage. (#4193)

* Add PushCSC for SparsePage.

* Move Push* definitions into cc file.
* Add std:: prefix to `size_t` make clang++ happy.
* Address monitor count == 0.

tests/cpp/data/test_data.cc

@@ -0,0 +1,55 @@
+#include <gtest/gtest.h>
+#include <vector>
+
+#include "xgboost/data.h"
+
+namespace xgboost {
+TEST(SparsePage, PushCSC) {
+  std::vector<size_t> offset {0};
+  std::vector<Entry> data;
+  SparsePage page;
+  page.offset.HostVector() = offset;
+  page.data.HostVector() = data;
+
+  offset = {0, 1, 4};
+  for (size_t i = 0; i < offset.back(); ++i) {
+    data.push_back(Entry(i, 0.1f));
+  }
+
+  SparsePage other;
+  other.offset.HostVector() = offset;
+  other.data.HostVector() = data;
+
+  page.PushCSC(other);
+
+  ASSERT_EQ(page.offset.HostVector().size(), offset.size());
+  ASSERT_EQ(page.data.HostVector().size(), data.size());
+  for (size_t i = 0; i < offset.size(); ++i) {
+    ASSERT_EQ(page.offset.HostVector()[i], offset[i]);
+  }
+  for (size_t i = 0; i < data.size(); ++i) {
+    ASSERT_EQ(page.data.HostVector()[i].index, data[i].index);
+  }
+
+  page.PushCSC(other);
+  ASSERT_EQ(page.offset.HostVector().size(), offset.size());
+  ASSERT_EQ(page.data.Size(), data.size() * 2);
+
+  for (size_t i = 0; i < offset.size(); ++i) {
+    ASSERT_EQ(page.offset.HostVector()[i], offset[i] * 2);
+  }
+
+  auto inst = page[0];
+  ASSERT_EQ(inst.size(), 2);
+  for (auto entry : inst) {
+    ASSERT_EQ(entry.index, 0);
+  }
+
+  inst = page[1];
+  ASSERT_EQ(inst.size(), 6);
+  std::vector<size_t> indices_sol {1, 2, 3};
+  for (size_t i = 0; i < inst.size(); ++i) {
+    ASSERT_EQ(inst[i].index, indices_sol[i % 3]);
+  }
+}
+}

tests/cpp/test_learner.cc

@@ -3,6 +3,7 @@
 #include <vector>
 #include "helpers.h"
 #include "xgboost/learner.h"
+#include "dmlc/filesystem.h"
 
 namespace xgboost {
 
@@ -92,4 +93,26 @@ TEST(Learner, CheckGroup) {
   delete pp_mat;
 }
 
+TEST(Learner, SLOW_CheckMultiBatch) {
+  using Arg = std::pair<std::string, std::string>;
+  // Create sufficiently large data to make two row pages
+  dmlc::TemporaryDirectory tempdir;
+  const std::string tmp_file = tempdir.path + "/big.libsvm";
+  CreateBigTestData(tmp_file, 5000000);
+  std::shared_ptr<DMatrix> dmat(xgboost::DMatrix::Load( tmp_file + "#" + tmp_file + ".cache", true, false));
+  EXPECT_TRUE(FileExists(tmp_file + ".cache.row.page"));
+  EXPECT_FALSE(dmat->SingleColBlock());
+  size_t num_row = dmat->Info().num_row_;
+  std::vector<bst_float> labels(num_row);
+  for (size_t i = 0; i < num_row; ++i) {
+    labels[i] = i % 2;
+  }
+  dmat->Info().SetInfo("label", labels.data(), DataType::kFloat32, num_row);
+  std::vector<std::shared_ptr<DMatrix>> mat{dmat};
+  auto learner = std::unique_ptr<Learner>(Learner::Create(mat));
+  learner->Configure({Arg{"objective", "binary:logistic"}});
+  learner->InitModel();
+  learner->UpdateOneIter(0, dmat.get());
+}
+
 }  // namespace xgboost