Cleanup set info. (#10139)

- Use the array interface internally.
- Deprecate `XGDMatrixSetDenseInfo`.
- Deprecate `XGDMatrixSetUIntInfo`.
- Move the handling of `DataType` into the deprecated C function.

---------

Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>

include/xgboost/c_api.h

@@ -1,5 +1,5 @@
 /**
- * Copyright 2015~2023 by XGBoost Contributors
+ * Copyright 2015-2024, XGBoost Contributors
  * \file c_api.h
  * \author Tianqi Chen
  * \brief C API of XGBoost, used for interfacing to other languages.
@@ -639,21 +639,14 @@ XGB_DLL int XGDMatrixSetInfoFromInterface(DMatrixHandle handle,
  * \param len length of array
  * \return 0 when success, -1 when failure happens
  */
-XGB_DLL int XGDMatrixSetFloatInfo(DMatrixHandle handle,
-                                  const char *field,
-                                  const float *array,
+XGB_DLL int XGDMatrixSetFloatInfo(DMatrixHandle handle, const char *field, const float *array,
                                   bst_ulong len);
-/*!
- * \brief set uint32 vector to a content in info
- * \param handle a instance of data matrix
- * \param field field name
- * \param array pointer to unsigned int vector
- * \param len length of array
- * \return 0 when success, -1 when failure happens
+/**
+ * @deprecated since 2.1.0
+ *
+ * Use @ref XGDMatrixSetInfoFromInterface instead.
  */
-XGB_DLL int XGDMatrixSetUIntInfo(DMatrixHandle handle,
-                                 const char *field,
-                                 const unsigned *array,
+XGB_DLL int XGDMatrixSetUIntInfo(DMatrixHandle handle, const char *field, const unsigned *array,
                                  bst_ulong len);
 
 /*!
@@ -725,42 +718,13 @@ XGB_DLL int XGDMatrixGetStrFeatureInfo(DMatrixHandle handle, const char *field,
                                        bst_ulong *size,
                                        const char ***out_features);
 
-/*!
- * \brief Set meta info from dense matrix.  Valid field names are:
+/**
+ * @deprecated since 2.1.0
  *
- *  - label
- *  - weight
- *  - base_margin
- *  - group
- *  - label_lower_bound
- *  - label_upper_bound
- *  - feature_weights
- *
- * \param handle An instance of data matrix
- * \param field  Field name
- * \param data   Pointer to consecutive memory storing data.
- * \param size   Size of the data, this is relative to size of type.  (Meaning NOT number
- *               of bytes.)
- * \param type   Indicator of data type.  This is defined in xgboost::DataType enum class.
- *    - float    = 1
- *    - double   = 2
- *    - uint32_t = 3
- *    - uint64_t = 4
- * \return 0 when success, -1 when failure happens
+ * Use @ref XGDMatrixSetInfoFromInterface instead.
  */
-XGB_DLL int XGDMatrixSetDenseInfo(DMatrixHandle handle, const char *field,
-                                  void const *data, bst_ulong size, int type);
-
-/*!
- * \brief (deprecated) Use XGDMatrixSetUIntInfo instead. Set group of the training matrix
- * \param handle a instance of data matrix
- * \param group pointer to group size
- * \param len length of array
- * \return 0 when success, -1 when failure happens
- */
-XGB_DLL int XGDMatrixSetGroup(DMatrixHandle handle,
-                              const unsigned *group,
-                              bst_ulong len);
+XGB_DLL int XGDMatrixSetDenseInfo(DMatrixHandle handle, const char *field, void const *data,
+                                  bst_ulong size, int type);
 
 /*!
  * \brief get float info vector from matrix.

include/xgboost/data.h

@@ -19,7 +19,6 @@
 #include <algorithm>
 #include <limits>
 #include <memory>
-#include <numeric>
 #include <string>
 #include <utility>
 #include <vector>
@@ -137,14 +136,6 @@ class MetaInfo {
    * \param fo The output stream.
    */
   void SaveBinary(dmlc::Stream* fo) const;
-  /*!
-   * \brief Set information in the meta info.
-   * \param key The key of the information.
-   * \param dptr The data pointer of the source array.
-   * \param dtype The type of the source data.
-   * \param num Number of elements in the source array.
-   */
-  void SetInfo(Context const& ctx, const char* key, const void* dptr, DataType dtype, size_t num);
   /*!
    * \brief Set information in the meta info with array interface.
    * \param key The key of the information.
@@ -517,10 +508,6 @@ class DMatrix {
   DMatrix()  = default;
   /*! \brief meta information of the dataset */
   virtual MetaInfo& Info() = 0;
-  virtual void SetInfo(const char* key, const void* dptr, DataType dtype, size_t num) {
-    auto const& ctx = *this->Ctx();
-    this->Info().SetInfo(ctx, key, dptr, dtype, num);
-  }
   virtual void SetInfo(const char* key, std::string const& interface_str) {
     auto const& ctx = *this->Ctx();
     this->Info().SetInfo(ctx, key, StringView{interface_str});

include/xgboost/linalg.h

@@ -190,13 +190,14 @@ constexpr auto ArrToTuple(T (&arr)[N]) {
 // uint division optimization inspired by the CIndexer in cupy.  Division operation is
 // slow on both CPU and GPU, especially 64 bit integer.  So here we first try to avoid 64
 // bit when the index is smaller, then try to avoid division when it's exp of 2.
-template <typename I, int32_t D>
+template <typename I, std::int32_t D>
 LINALG_HD auto UnravelImpl(I idx, common::Span<size_t const, D> shape) {
-  size_t index[D]{0};
+  std::size_t index[D]{0};
   static_assert(std::is_signed<decltype(D)>::value,
                 "Don't change the type without changing the for loop.");
+  auto const sptr = shape.data();
   for (int32_t dim = D; --dim > 0;) {
-    auto s = static_cast<std::remove_const_t<std::remove_reference_t<I>>>(shape[dim]);
+    auto s = static_cast<std::remove_const_t<std::remove_reference_t<I>>>(sptr[dim]);
     if (s & (s - 1)) {
       auto t = idx / s;
       index[dim] = idx - t * s;
@@ -745,6 +746,14 @@ auto ArrayInterfaceStr(TensorView<T, D> const &t) {
   return str;
 }
 
+template <typename T>
+auto Make1dInterface(T const *vec, std::size_t len) {
+  Context ctx;
+  auto t = linalg::MakeTensorView(&ctx, common::Span{vec, len}, len);
+  auto str = linalg::ArrayInterfaceStr(t);
+  return str;
+}
+
 /**
  * \brief A tensor storage. To use it for other functionality like slicing one needs to
  *        obtain a view first.  This way we can use it on both host and device.

include/xgboost/span.h

@@ -30,9 +30,8 @@
 #define XGBOOST_SPAN_H_
 
 #include <xgboost/base.h>
-#include <xgboost/logging.h>
 
-#include <cinttypes>  // size_t
+#include <cstddef>  // size_t
 #include <cstdio>
 #include <iterator>
 #include <limits>  // numeric_limits
@@ -73,8 +72,7 @@
 
 #endif  // defined(_MSC_VER) && _MSC_VER < 1910
 
-namespace xgboost {
-namespace common {
+namespace xgboost::common {
 
 #if defined(__CUDA_ARCH__)
 // Usual logging facility is not available inside device code.
@@ -707,8 +705,8 @@ class IterSpan {
     return it_ + size();
   }
 };
-}  // namespace common
-}  // namespace xgboost
+}  // namespace xgboost::common
+
 
 #if defined(_MSC_VER) &&_MSC_VER < 1910
 #undef constexpr