Extract string view. (#7416)

* Add equality operators.
* Return a view in substr.
* Add proper iterator types.

include/xgboost/intrusive_ptr.h

@@ -9,6 +9,7 @@
 #include <atomic>
 #include <cinttypes>
 #include <functional>
+#include <ostream>
 
 namespace xgboost {
 /*!

include/xgboost/json.h

@@ -4,16 +4,17 @@
 #ifndef XGBOOST_JSON_H_
 #define XGBOOST_JSON_H_
 
+#include <xgboost/intrusive_ptr.h>
 #include <xgboost/logging.h>
 #include <xgboost/parameter.h>
-#include <xgboost/intrusive_ptr.h>
+#include <xgboost/string_view.h>
 
+#include <functional>
 #include <map>
 #include <memory>
-#include <vector>
-#include <functional>
-#include <utility>
 #include <string>
+#include <utility>
+#include <vector>
 
 namespace xgboost {
 
@@ -298,43 +299,6 @@ class JsonBoolean : public Value {
   }
 };
 
-struct StringView {
- private:
-  using CharT = char;  // unsigned char
-  using Traits = std::char_traits<CharT>;
-  CharT const* str_;
-  size_t size_;
-
- public:
-  StringView() = default;
-  StringView(CharT const* str, size_t size) : str_{str}, size_{size} {}
-  explicit StringView(std::string const& str): str_{str.c_str()}, size_{str.size()} {}
-  explicit StringView(CharT const* str) : str_{str}, size_{Traits::length(str)} {}
-
-  CharT const& operator[](size_t p) const { return str_[p]; }
-  CharT const& at(size_t p) const {  // NOLINT
-    CHECK_LT(p, size_);
-    return str_[p];
-  }
-  size_t size() const { return size_; }  // NOLINT
-  // Copies a portion of string.  Since we don't have std::from_chars and friends here, so
-  // copying substring is necessary for appending `\0`.  It's not too bad since string by
-  // default has small vector optimization, which is enabled by most if not all modern
-  // compilers for numeric values.
-  std::string substr(size_t beg, size_t n) const {  // NOLINT
-    CHECK_LE(beg, size_);
-    return std::string {str_ + beg, n < (size_ - beg) ? n : (size_ - beg)};
-  }
-  CharT const* c_str() const { return str_; }  // NOLINT
-
-  CharT const* cbegin() const { return str_; }         // NOLINT
-  CharT const* cend() const { return str_ + size(); }  // NOLINT
-  CharT const* begin() const { return str_; }          // NOLINT
-  CharT const* end() const { return str_ + size(); }   // NOLINT
-};
-
-std::ostream &operator<<(std::ostream &os, StringView const v);
-
 /*!
  * \brief Data structure representing JSON format.
  *
@@ -451,7 +415,7 @@ class Json {
 };
 
 template <typename T>
-bool IsA(Json const j) {
+bool IsA(Json const& j) {
   auto const& v = j.GetValue();
   return IsA<T>(&v);
 }

include/xgboost/json_io.h

@@ -84,12 +84,12 @@ class JsonReader {
     std::string msg = "Expecting: \"";
     msg += c;
     msg += "\", got: \"";
-    if (got == -1) {
+    if (got == EOF) {
       msg += "EOF\"";
     } else if (got == 0) {
       msg += "\\0\"";
     } else {
-      msg += std::to_string(got) + " \"";
+      msg += (got <= 127 ? std::string{got} : std::to_string(got)) + " \"";  // NOLINT
     }
     Error(msg);
   }

include/xgboost/string_view.h

@@ -0,0 +1,81 @@
+/*!
+ * Copyright 2021 by XGBoost Contributors
+ */
+#ifndef XGBOOST_STRING_VIEW_H_
+#define XGBOOST_STRING_VIEW_H_
+#include <xgboost/logging.h>
+
+#include <algorithm>
+#include <iterator>
+#include <ostream>
+#include <string>
+
+namespace xgboost {
+struct StringView {
+ private:
+  using CharT = char;  // unsigned char
+  using Traits = std::char_traits<CharT>;
+  CharT const* str_{nullptr};
+  size_t size_{0};
+
+ public:
+  using iterator = const CharT*;                                   // NOLINT
+  using const_iterator = iterator;                                 // NOLINT
+  using reverse_iterator = std::reverse_iterator<const_iterator>;  // NOLINT
+  using const_reverse_iterator = reverse_iterator;                 // NOLINT
+
+ public:
+  constexpr StringView() = default;
+  constexpr StringView(CharT const* str, size_t size) : str_{str}, size_{size} {}
+  explicit StringView(std::string const& str) : str_{str.c_str()}, size_{str.size()} {}
+  StringView(CharT const* str) : str_{str}, size_{Traits::length(str)} {}  // NOLINT
+
+  CharT const& operator[](size_t p) const { return str_[p]; }
+  CharT const& at(size_t p) const {  // NOLINT
+    CHECK_LT(p, size_);
+    return str_[p];
+  }
+  constexpr size_t size() const { return size_; }  // NOLINT
+  StringView substr(size_t beg, size_t n) const {  // NOLINT
+    CHECK_LE(beg, size_);
+    size_t len = std::min(n, size_ - beg);
+    return {str_ + beg, len};
+  }
+  CharT const* c_str() const { return str_; }  // NOLINT
+
+  constexpr CharT const* cbegin() const { return str_; }         // NOLINT
+  constexpr CharT const* cend() const { return str_ + size(); }  // NOLINT
+  constexpr CharT const* begin() const { return str_; }          // NOLINT
+  constexpr CharT const* end() const { return str_ + size(); }   // NOLINT
+
+  const_reverse_iterator rbegin() const noexcept {  // NOLINT
+    return const_reverse_iterator(this->end());
+  }
+  const_reverse_iterator crbegin() const noexcept {  // NOLINT
+    return const_reverse_iterator(this->end());
+  }
+  const_reverse_iterator rend() const noexcept {  // NOLINT
+    return const_reverse_iterator(this->begin());
+  }
+  const_reverse_iterator crend() const noexcept {  // NOLINT
+    return const_reverse_iterator(this->begin());
+  }
+};
+
+inline std::ostream& operator<<(std::ostream& os, StringView const v) {
+  for (auto c : v) {
+    os.put(c);
+  }
+  return os;
+}
+
+inline bool operator==(StringView l, StringView r) {
+  if (l.size() != r.size()) {
+    return false;
+  }
+  return std::equal(l.cbegin(), l.cend(), r.cbegin());
+}
+
+inline bool operator!=(StringView l, StringView r) { return !(l == r); }
+}  // namespace xgboost
+#endif  // XGBOOST_STRING_VIEW_H_