[BLOCKING] Handle empty rows in data iterators correctly (#5929)

* [jvm-packages] Handle empty rows in data iterators correctly

* Fix clang-tidy error

* last empty row

* Add comments [skip ci]

Co-authored-by: Nan Zhu <nanzhu@uber.com>

jvm-packages/xgboost4j-spark/src/test/scala/ml/dmlc/xgboost4j/scala/spark/MissingValueHandlingSuite.scala

@@ -171,4 +171,31 @@ class MissingValueHandlingSuite extends FunSuite with PerTest {
     val model = new XGBoostClassifier(paramMap).fit(inputDF)
     model.transform(inputDF).collect()
   }
+
+  // https://github.com/dmlc/xgboost/pull/5929
+  test("handle the empty last row correctly with a missing value as 0") {
+    val spark = ss
+    import spark.implicits._
+    // spark uses 1.5 * (nnz + 1.0) < size as the condition to decide whether using sparse or dense
+    // vector,
+    val testDF = Seq(
+      (7.0f, 0.0f, -1.0f, 1.0f, 1.0),
+      (1.0f, 0.0f, 1.0f, 1.0f, 1.0),
+      (0.0f, 1.0f, 0.0f, 1.0f, 0.0),
+      (1.0f, 0.0f, 1.0f, 1.0f, 1.0),
+      (1.0f, -1.0f, 0.0f, 1.0f, 0.0),
+      (0.0f, 0.0f, 0.0f, 1.0f, 1.0),
+      (0.0f, 0.0f, 0.0f, 0.0f, 0.0)
+    ).toDF("col1", "col2", "col3", "col4", "label")
+    val vectorAssembler = new VectorAssembler()
+      .setInputCols(Array("col1", "col2", "col3", "col4"))
+      .setOutputCol("features")
+    val inputDF = vectorAssembler.transform(testDF).select("features", "label")
+    inputDF.show()
+    val paramMap = List("eta" -> "1", "max_depth" -> "2",
+      "objective" -> "binary:logistic", "missing" -> 0.0f,
+      "num_workers" -> 1, "allow_non_zero_for_missing" -> "true").toMap
+    val model = new XGBoostClassifier(paramMap).fit(inputDF)
+    model.transform(inputDF).collect()
+  }
 }

src/data/data.cc

@@ -833,9 +833,9 @@ uint64_t SparsePage::Push(const AdapterBatchT& batch, float missing, int nthread
   uint64_t max_columns = 0;
 
   // First-pass over the batch counting valid elements
-  size_t num_lines = batch.Size();
+  size_t batch_size = batch.Size();
 #pragma omp parallel for schedule(static)
-  for (omp_ulong i = 0; i < static_cast<omp_ulong>(num_lines);
+  for (omp_ulong i = 0; i < static_cast<omp_ulong>(batch_size);
        ++i) {  // NOLINT(*)
     int tid = omp_get_thread_num();
     auto line = batch.GetLine(i);
@@ -856,7 +856,7 @@ uint64_t SparsePage::Push(const AdapterBatchT& batch, float missing, int nthread
 
   // Second pass over batch, placing elements in correct position
 #pragma omp parallel for schedule(static)
-  for (omp_ulong i = 0; i < static_cast<omp_ulong>(num_lines);
+  for (omp_ulong i = 0; i < static_cast<omp_ulong>(batch_size);
        ++i) {  // NOLINT(*)
     int tid = omp_get_thread_num();
     auto line = batch.GetLine(i);

src/data/simple_dmatrix.cc

@@ -6,6 +6,7 @@
  */
 #include <vector>
 #include <limits>
+#include <type_traits>
 #include <algorithm>
 
 #include "xgboost/data.h"
@@ -103,6 +104,8 @@ SimpleDMatrix::SimpleDMatrix(AdapterT* adapter, float missing, int nthread) {
   auto& offset_vec = sparse_page_.offset.HostVector();
   auto& data_vec = sparse_page_.data.HostVector();
   uint64_t inferred_num_columns = 0;
+  uint64_t total_batch_size = 0;
+    // batch_size is either number of rows or cols, depending on data layout
 
   adapter->BeforeFirst();
   // Iterate over batches of input data
@@ -110,6 +113,7 @@ SimpleDMatrix::SimpleDMatrix(AdapterT* adapter, float missing, int nthread) {
     auto& batch = adapter->Value();
     auto batch_max_columns = sparse_page_.Push(batch, missing, nthread);
     inferred_num_columns = std::max(batch_max_columns, inferred_num_columns);
+    total_batch_size += batch.Size();
     // Append meta information if available
     if (batch.Labels() != nullptr) {
       auto& labels = info_.labels_.HostVector();
@@ -153,16 +157,30 @@ SimpleDMatrix::SimpleDMatrix(AdapterT* adapter, float missing, int nthread) {
     info_.num_col_ = adapter->NumColumns();
   }
 
+
   // Synchronise worker columns
   rabit::Allreduce<rabit::op::Max>(&info_.num_col_, 1);
 
   if (adapter->NumRows() == kAdapterUnknownSize) {
+    using IteratorAdapterT
+      = IteratorAdapter<DataIterHandle, XGBCallbackDataIterNext, XGBoostBatchCSR>;
+    // If AdapterT is either IteratorAdapter or FileAdapter type, use the total batch size to
+    // determine the correct number of rows, as offset_vec may be too short
+    if (std::is_same<AdapterT, IteratorAdapterT>::value
+        || std::is_same<AdapterT, FileAdapter>::value) {
+      info_.num_row_ = total_batch_size;
+      // Ensure offset_vec.size() - 1 == [number of rows]
+      while (offset_vec.size() - 1 < total_batch_size) {
+        offset_vec.emplace_back(offset_vec.back());
+      }
+    } else {
+      CHECK((std::is_same<AdapterT, CSCAdapter>::value)) << "Expecting CSCAdapter";
       info_.num_row_ = offset_vec.size() - 1;
+    }
   } else {
     if (offset_vec.empty()) {
       offset_vec.emplace_back(0);
     }
-
     while (offset_vec.size() - 1 < adapter->NumRows()) {
       offset_vec.emplace_back(offset_vec.back());
     }

tests/cpp/data/test_adapter.cc

@@ -26,12 +26,13 @@ TEST(Adapter, CSRAdapter) {
   EXPECT_EQ(line0.GetElement(1).value, 2);
 
   auto line1 = batch.GetLine(1);
-  EXPECT_EQ(line1 .GetElement(0).value, 3);
+  EXPECT_EQ(line1.GetElement(0).value, 3);
-  EXPECT_EQ(line1 .GetElement(1).value, 4);
+  EXPECT_EQ(line1.GetElement(1).value, 4);
+
   auto line2 = batch.GetLine(2);
-  EXPECT_EQ(line2 .GetElement(0).value, 5);
+  EXPECT_EQ(line2.GetElement(0).value, 5);
-  EXPECT_EQ(line2 .GetElement(0).row_idx, 2);
+  EXPECT_EQ(line2.GetElement(0).row_idx, 2);
-  EXPECT_EQ(line2 .GetElement(0).column_idx, 1);
+  EXPECT_EQ(line2.GetElement(0).column_idx, 1);
 }
 
 TEST(Adapter, CSCAdapterColsMoreThanRows) {
@@ -73,10 +74,11 @@ class CSRIterForTest {
   std::vector<std::remove_pointer<decltype(std::declval<XGBoostBatchCSR>().index)>::type>
       feature_idx_ {0, 1, 0, 1, 1};
   std::vector<std::remove_pointer<decltype(std::declval<XGBoostBatchCSR>().offset)>::type>
-      row_ptr_ {0, 2, 4, 5};
+      row_ptr_ {0, 2, 4, 5, 5};
   size_t iter_ {0};
 
  public:
+  size_t static constexpr kRows { 4 };  // Test for the last row being empty
   size_t static constexpr kCols { 13 };  // Test for having some missing columns
 
   XGBoostBatchCSR Next() {
@@ -88,7 +90,7 @@ class CSRIterForTest {
     batch.offset = dmlc::BeginPtr(row_ptr_);
     batch.index = dmlc::BeginPtr(feature_idx_);
     batch.value = dmlc::BeginPtr(data_);
-    batch.size = 3;
+    batch.size = kRows;
 
     batch.label = nullptr;
     batch.weight = nullptr;
@@ -117,16 +119,23 @@ int CSRSetDataNextForTest(DataIterHandle data_handle,
   }
 }
 
-TEST(Adapter, IteratorAdaper) {
+TEST(Adapter, IteratorAdapter) {
   CSRIterForTest iter;
   data::IteratorAdapter<DataIterHandle, XGBCallbackDataIterNext,
                         XGBoostBatchCSR> adapter{&iter, CSRSetDataNextForTest};
-  constexpr size_t kRows { 6 };
+  constexpr size_t kRows { 8 };
 
   std::unique_ptr<DMatrix> data {
     DMatrix::Create(&adapter, std::numeric_limits<float>::quiet_NaN(), 1)
   };
   ASSERT_EQ(data->Info().num_col_, CSRIterForTest::kCols);
   ASSERT_EQ(data->Info().num_row_, kRows);
+  int num_batch = 0;
+  for (auto const& batch : data->GetBatches<SparsePage>()) {
+    ASSERT_EQ(batch.offset.HostVector(), std::vector<bst_row_t>({0, 2, 4, 5, 5, 7, 9, 10, 10}));
+    ++num_batch;
+  }
+  ASSERT_EQ(num_batch, 1);
 }
+
 }  // namespace xgboost

tests/cpp/data/test_simple_dmatrix.cc

@@ -185,16 +185,22 @@ TEST(SimpleDMatrix, FromCSC) {
 TEST(SimpleDMatrix, FromFile) {
   std::string filename = "test.libsvm";
   CreateBigTestData(filename, 3 * 5);
+  // Add an empty row at the end of the matrix
+  {
+    std::ofstream fo(filename, std::ios::app | std::ios::out);
+    fo << "0\n";
+  }
+  constexpr size_t kExpectedNumRow = 6;
   std::unique_ptr<dmlc::Parser<uint32_t>> parser(
       dmlc::Parser<uint32_t>::Create(filename.c_str(), 0, 1, "auto"));
 
-  auto verify_batch = [](SparsePage const &batch) {
+  auto verify_batch = [kExpectedNumRow](SparsePage const &batch) {
-    EXPECT_EQ(batch.Size(), 5);
+    EXPECT_EQ(batch.Size(), kExpectedNumRow);
     EXPECT_EQ(batch.offset.HostVector(),
-              std::vector<bst_row_t>({0, 3, 6, 9, 12, 15}));
+              std::vector<bst_row_t>({0, 3, 6, 9, 12, 15, 15}));
     EXPECT_EQ(batch.base_rowid, 0);
 
-    for (auto i = 0ull; i < batch.Size(); i++) {
+    for (auto i = 0ull; i < batch.Size() - 1; i++) {
       if (i % 2 == 0) {
         EXPECT_EQ(batch[i][0].index, 0);
         EXPECT_EQ(batch[i][1].index, 1);