Fix spelling in documents (#6948)

* Update roxygen2 doc.

Co-authored-by: fis <jm.yuan@outlook.com>

src/common/charconv.cc

@@ -1,7 +1,7 @@
 /*!
  * Copyright 2020 by XGBoost Contributors
  *
- * \brief An implemenation of Ryu algorithm:
+ * \brief An implementation of Ryu algorithm:
  *
  * https://dl.acm.org/citation.cfm?id=3192369
  *
@@ -686,7 +686,7 @@ int32_t ToCharsFloatImpl(float f, char * const result) {
 
 // This is an implementation for base 10 inspired by the one in libstdc++v3.  The general
 // scheme is by decomposing the value into multiple combination of base (which is 10) by
-// mod, until the value is lesser than 10, then last char is just char '0' (ascii 48) plus
+// mod, until the value is lesser than 10, then last char is just char '0' (ASCII 48) plus
 // that value.  Other popular implementations can be found in RapidJson and libc++ (in
 // llvm-project), which uses the same general work flow with the same look up table, but
 // probably with better performance as they are more complicated.

src/common/common.h

@@ -55,7 +55,7 @@ namespace xgboost {
 namespace common {
 /*!
  * \brief Split a string by delimiter
- * \param s String to be splitted.
+ * \param s String to be split.
  * \param delim The delimiter.
  */
 inline std::vector<std::string> Split(const std::string& s, char delim) {

src/common/device_helpers.cuh

@@ -253,7 +253,7 @@ __global__ void LaunchNKernel(int device_idx, size_t begin, size_t end,
  *   function as argument.  Hence functions like `LaunchN` cannot use this wrapper.
  *
  * - With c++ initialization list `{}` syntax, you are forced to comply with the CUDA type
- *   spcification.
+ *   specification.
  */
 class LaunchKernel {
   size_t shmem_size_;
@@ -930,7 +930,7 @@ class SegmentSorter {
   // Items sorted within the group
   caching_device_vector<T> ditems_;
 
-  // Original position of the items before they are sorted descendingly within its groups
+  // Original position of the items before they are sorted descending within their groups
   caching_device_vector<uint32_t> doriginal_pos_;
 
   // Segments within the original list that delineates the different groups

src/common/hist_util.h

@@ -81,7 +81,7 @@ class HistogramCuts {
   }
 
   // Getters.  Cuts should be of no use after building histogram indices, but currently
-  // it's deeply linked with quantile_hist, gpu sketcher and gpu_hist.  So we preserve
+  // they are deeply linked with quantile_hist, gpu sketcher and gpu_hist, so we preserve
   // these for now.
   std::vector<uint32_t> const& Ptrs()      const { return cut_ptrs_.ConstHostVector();   }
   std::vector<float>    const& Values()    const { return cut_values_.ConstHostVector(); }
@@ -247,7 +247,7 @@ struct GHistIndexMatrix {
   // Create a global histogram matrix, given cut
   void Init(DMatrix* p_fmat, int max_num_bins);
 
-  // specific method for sparse data as no posibility to reduce allocated memory
+  // specific method for sparse data as no possibility to reduce allocated memory
   template <typename BinIdxType, typename GetOffset>
   void SetIndexData(common::Span<BinIdxType> index_data_span,
                     size_t batch_threads, const SparsePage &batch,

src/common/json.cc

@@ -394,7 +394,7 @@ Json JsonReader::Parse() {
       return ParseArray();
     } else if ( c == '-' || std::isdigit(c) ||
                 c == 'N' || c == 'I') {
-      // For now we only accept `NaN`, not `nan` as the later violiates LR(1) with `null`.
+      // For now we only accept `NaN`, not `nan` as the later violates LR(1) with `null`.
       return ParseNumber();
     } else if ( c == '\"' ) {
       return ParseString();

src/common/math.h

@@ -77,7 +77,7 @@ XGBOOST_DEVICE inline void Softmax(Iterator start, Iterator end) {
 
 /*!
  * \brief Find the maximum iterator within the iterators
- * \param begin The begining iterator.
+ * \param begin The beginning iterator.
  * \param end The end iterator.
  * \return the iterator point to the maximum value.
  * \tparam Iterator The type of the iterator.
@@ -107,7 +107,7 @@ inline float LogSum(float x, float y) {
 
 /*!
  * \brief perform numerically safe logsum
- * \param begin The begining iterator.
+ * \param begin The beginning iterator.
  * \param end The end iterator.
  * \return the iterator point to the maximum value.
  * \tparam Iterator The type of the iterator.
@@ -135,7 +135,7 @@ inline static bool CmpSecond(const std::pair<float, unsigned> &a,
   return a.second > b.second;
 }
 
-// Redefined here to workaround a VC bug that doesn't support overloadng for integer
+// Redefined here to workaround a VC bug that doesn't support overloading for integer
 // types.
 template <typename T>
 XGBOOST_DEVICE typename std::enable_if<

src/common/quantile.cc

@@ -55,7 +55,7 @@ HostSketchContainer::CalcColumnSize(SparsePage const &batch,
 std::vector<bst_feature_t> HostSketchContainer::LoadBalance(
     SparsePage const &batch, bst_feature_t n_columns, size_t const nthreads) {
   /* Some sparse datasets have their mass concentrating on small number of features.  To
-   * avoid wating for a few threads running forever, we here distirbute different number
+   * avoid waiting for a few threads running forever, we here distribute different number
    * of columns to different threads according to number of entries.
    */
   auto page = batch.GetView();

src/common/quantile.cu

@@ -184,9 +184,9 @@ common::Span<thrust::tuple<uint64_t, uint64_t>> MergePath(
       });
 
   // Compute the index for both x and y (which of the element in a and b are used in each
-  // comparison) by scaning the binary merge path.  Take output [(x_0, y_0), (x_0, y_1),
+  // comparison) by scanning the binary merge path.  Take output [(x_0, y_0), (x_0, y_1),
   // ...] as an example, the comparison between (x_0, y_0) adds 1 step in the merge path.
-  // Asumming y_0 is less than x_0 so this step is torward the end of y.  After the
+  // Assuming y_0 is less than x_0 so this step is toward the end of y.  After the
   // comparison, index of y is incremented by 1 from y_0 to y_1, and at the same time, y_0
   // is landed into output as the first element in merge result.  The scan result is the
   // subscript of x and y.
@@ -367,7 +367,7 @@ void SketchContainer::Push(Span<Entry const> entries, Span<size_t> columns_ptr,
 size_t SketchContainer::ScanInput(Span<SketchEntry> entries, Span<OffsetT> d_columns_ptr_in) {
   /* There are 2 types of duplication.  First is duplicated feature values, which comes
    * from user input data.  Second is duplicated sketching entries, which is generated by
-   * prunning or merging. We preserve the first type and remove the second type.
+   * pruning or merging. We preserve the first type and remove the second type.
    */
   timer_.Start(__func__);
   dh::safe_cuda(cudaSetDevice(device_));

src/common/survival_util.h

@@ -44,7 +44,7 @@ constexpr double kMaxGradient = 15.0;
 constexpr double kMinHessian = 1e-16;  // Ensure that no data point gets zero hessian
 constexpr double kMaxHessian = 15.0;
 
-constexpr double kEps = 1e-12;  // A denomitor in a fraction should not be too small
+constexpr double kEps = 1e-12;  // A denominator in a fraction should not be too small
 
 // Clip (limit) x to fit range [x_min, x_max].
 // If x < x_min, return x_min; if x > x_max, return x_max; if x_min <= x <= x_max, return x.

src/common/transform.h

@@ -52,7 +52,7 @@ __global__ void LaunchCUDAKernel(Functor _func, Range _range,
  *
  *     If you use it in a function that can be compiled by both nvcc and host
  *     compiler, the behaviour is un-defined!  Because your function is NOT
- *     duplicated by `CompiledWithCuda`. At link time, cuda compiler resolution
+ *     duplicated by `CompiledWithCuda`. At link time, CUDA compiler resolution
  *     will merge functions with same signature.
  */
 template <bool CompiledWithCuda = WITH_CUDA()>
@@ -155,7 +155,7 @@ class Transform {
           _func, shard_range, UnpackHDVOnDevice(_vectors)...);
     }
 #else
-    /*! \brief Dummy funtion defined when compiling for CPU.  */
+    /*! \brief Dummy function defined when compiling for CPU.  */
     template <typename std::enable_if<!CompiledWithCuda>::type* = nullptr,
               typename... HDV>
     void LaunchCUDA(Functor _func, HDV*...) const {

src/common/version.cc

@@ -36,7 +36,7 @@ Version::TripletT Version::Load(Json const& in) {
 
 Version::TripletT Version::Load(dmlc::Stream* fi) {
   XGBoostVersionT major{0}, minor{0}, patch{0};
-  // This is only used in DMatrix serialization, so doesn't break model compability.
+  // This is only used in DMatrix serialization, so doesn't break model compatibility.
   std::string msg { "Incorrect version format found in binary file.  "
                     "Binary file from XGBoost < 1.0.0 is no longer supported. "
                     "Please generate it again." };

src/common/version.h

@@ -17,7 +17,7 @@ struct Version {
   using TripletT = std::tuple<XGBoostVersionT, XGBoostVersionT, XGBoostVersionT>;
   static const TripletT kInvalid;
 
-  // Save/Load version info to Json document
+  // Save/Load version info to JSON document
   static TripletT Load(Json const& in);
   static void Save(Json* out);