format.h

Go to the documentation of this file.

 
/******************************************************************************
The MIT License(MIT)
 
Embedded Template Library.
https://github.com/ETLCPP/etl
https://www.etlcpp.com
 
Copyright(c) 2025 BMW AG
 
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files(the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions :
 
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
 
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
******************************************************************************/
 
#ifndef ETL_FORMAT_INCLUDED
#define ETL_FORMAT_INCLUDED
 
#include "platform.h"
 
#include "algorithm.h"
#include "array.h"
#include "array_view.h"
#include "error_handler.h"
#include "limits.h"
#include "math.h"
#include "optional.h"
#include "span.h"
#include "string.h"
#include "string_view.h"
#include "type_traits.h"
#include "utility.h"
#include "variant.h"
#include "visitor.h"
 
#if ETL_USING_FORMAT_FLOATING_POINT
  #include <cmath>
#endif
 
#if ETL_USING_CPP11
 
namespace etl
{
  class format_exception : public etl::exception
  {
  public:
 
    format_exception(string_type reason_, string_type file_name_, numeric_type line_number_)
      : exception(reason_, file_name_, line_number_)
    {
    }
  };
 
  class bad_format_string_exception : public etl::format_exception
  {
  public:
 
    bad_format_string_exception(string_type file_name_, numeric_type line_number_)
      : etl::format_exception(ETL_ERROR_TEXT("format:bad", ETL_FORMAT_FILE_ID"A"), file_name_, line_number_)
    {
    }
  };
 
  #if ETL_USING_CPP20
  namespace private_format_check
  {
    // Type category for compile-time type/specifier compatibility checking
    enum class type_category
    {
      NONE,    // monostate
      BOOLEAN, // bool
      CHAR,    // char
      INTEGER, // int, unsigned, long long, unsigned long long, short, etc.
      FLOAT,   // float, double, long double
      STRING,  // const char*, string_view
      POINTER  // const void*
    };
 
    // Map a type to its category. Decays and removes cv-qualifiers.
    template <class T>
    constexpr type_category get_type_category()
    {
      using U = typename etl::remove_cv<typename etl::remove_reference<T>::type>::type;
 
      // Order matters: bool before integral, char before integral
      if (etl::is_same<U, bool>::value)
        return type_category::BOOLEAN;
      if (etl::is_same<U, char>::value)
        return type_category::CHAR;
      if (etl::is_same<U, signed char>::value)
        return type_category::CHAR;
      if (etl::is_same<U, unsigned char>::value)
        return type_category::CHAR;
      if (etl::is_integral<U>::value)
        return type_category::INTEGER;
      if (etl::is_same<U, float>::value)
        return type_category::FLOAT;
      if (etl::is_same<U, double>::value)
        return type_category::FLOAT;
      if (etl::is_same<U, long double>::value)
        return type_category::FLOAT;
      if (etl::is_same<U, const char*>::value)
        return type_category::STRING;
      if (etl::is_same<U, char*>::value)
        return type_category::STRING;
      if (etl::is_same<U, etl::string_view>::value)
        return type_category::STRING;
      if (etl::is_base_of<etl::istring, U>::value)
        return type_category::STRING;
      if (etl::is_pointer<U>::value)
        return type_category::POINTER;
      if (etl::is_same<U, const void*>::value)
        return type_category::POINTER;
      if (etl::is_same<U, void*>::value)
        return type_category::POINTER;
      return type_category::NONE; // unknown type: custom formatter, be permissive
    }
 
    // Check if a format type character is valid for a given type category.
    // '\0' means no explicit type was specified (always valid — uses default presentation).
    inline constexpr bool ct_check_type_spec(type_category cat, char type_char)
    {
      if (type_char == '\0')
      {
        return true; // no explicit type: always OK, uses default
      }
 
      switch (cat)
      {
        case type_category::BOOLEAN:
          // bool: s (as "true"/"false"), b, B, c, d, o, x, X (as integer 0/1)
          return type_char == 's' || type_char == 'b' || type_char == 'B' || type_char == 'c' || type_char == 'd' || type_char == 'o'
                 || type_char == 'x' || type_char == 'X';
 
        case type_category::CHAR:
          // char: c (default), b, B, d, o, x, X (as integer), s, ?
          return type_char == 'c' || type_char == '?' || type_char == 'b' || type_char == 'B' || type_char == 'd' || type_char == 'o'
                 || type_char == 'x' || type_char == 'X' || type_char == 's';
 
        case type_category::INTEGER:
          // integers: b, B, c, d, o, x, X
          return type_char == 'b' || type_char == 'B' || type_char == 'c' || type_char == 'd' || type_char == 'o' || type_char == 'x'
                 || type_char == 'X';
 
        case type_category::FLOAT:
          // floats: a, A, e, E, f, F, g, G
          return type_char == 'a' || type_char == 'A' || type_char == 'e' || type_char == 'E' || type_char == 'f' || type_char == 'F'
                 || type_char == 'g' || type_char == 'G';
 
        case type_category::STRING:
          // strings: s, ?
          return type_char == 's' || type_char == '?';
 
        case type_category::POINTER:
          // pointers: p, P
          return type_char == 'p' || type_char == 'P';
 
        case type_category::NONE:
        default: return true; // unknown/custom type: be permissive, let runtime handle it
      }
    }
 
    inline constexpr bool ct_is_digit(char c)
    {
      return c >= '0' && c <= '9';
    }
 
    // Parse an unsigned integer from fmt starting at pos. Updates pos past the digits.
    // Returns the parsed number, or -1 if no digits found, or -2 on overflow.
    inline constexpr int ct_parse_num(const char* fmt, int& pos)
    {
      if (!ct_is_digit(fmt[pos]))
      {
        return -1;
      }
      int result = 0;
      while (ct_is_digit(fmt[pos]))
      {
        int new_result = result * 10 + (fmt[pos] - '0');
        if (new_result < result)
        {
          // Overflow detected
          return -2;
        }
        result = new_result;
        ++pos;
      }
      return result;
    }
 
    inline constexpr bool ct_is_align(char c)
    {
      return c == '<' || c == '>' || c == '^';
    }
 
    inline constexpr bool ct_is_sign(char c)
    {
      return c == '+' || c == '-' || c == ' ';
    }
 
    inline constexpr bool ct_is_type(char c)
    {
      // All valid type characters from the format spec
      return (c == 's') || (c == '?') || (c == 'b') || (c == 'B') || (c == 'c') || (c == 'd') || (c == 'o') || (c == 'x') || (c == 'X') || (c == 'a')
             || (c == 'A') || (c == 'e') || (c == 'E') || (c == 'f') || (c == 'F') || (c == 'g') || (c == 'G') || (c == 'p') || (c == 'P');
    }
 
    // Validate a nested replacement field like {}, {0}, {1} inside width/precision.
    // pos should be at the '{'. Updates pos past the closing '}'.
    // Updates auto_count / has_manual / has_auto. Returns false on error.
    inline constexpr bool ct_parse_nested_replacement(const char* fmt, int& pos, int n_args, int& auto_count, bool& has_auto, bool& has_manual)
    {
      if (fmt[pos] != '{')
        return false;
      ++pos; // skip '{'
 
      int num = ct_parse_num(fmt, pos);
      if (num == -2)
        return false; // overflow
      if (num >= 0)
      {
        // manual index
        if (has_auto)
          return false; // mixing
        has_manual = true;
        if (num >= n_args)
          return false;
      }
      else
      {
        // automatic index
        if (has_manual)
          return false; // mixing
        has_auto = true;
        if (auto_count >= n_args)
          return false;
        ++auto_count;
      }
 
      if (fmt[pos] != '}')
        return false;
      ++pos; // skip '}'
      return true;
    }
 
    // Skip/validate the format-spec portion after the colon:
    //   [[fill]align][sign][#][0][width][.precision][L][type]
    // pos is right after ':'. Returns false on invalid spec.
    // parsed_type is set to the type character found, or '\0' if none.
    inline constexpr bool ct_skip_format_spec(const char* fmt, int& pos, int n_args, int& auto_count, bool& has_auto, bool& has_manual,
                                              char& parsed_type, bool& parsed_has_precision)
    {
      parsed_type          = '\0';
      parsed_has_precision = false;
 
      if (fmt[pos] == '\0' || fmt[pos] == '}')
      {
        return true; // empty spec is valid
      }
 
      // fill-and-align: either [align] or [fill][align]
      // Look ahead: if second char is an align char, first is fill
      if (fmt[pos + 1] != '\0' && ct_is_align(fmt[pos + 1]))
      {
        char fill = fmt[pos];
        if (fill == '{' || fill == '}')
          return false; // { and } not allowed as fill
        pos += 2;       // skip fill + align
      }
      else if (ct_is_align(fmt[pos]))
      {
        ++pos; // skip align only
      }
 
      // sign
      if (ct_is_sign(fmt[pos]))
      {
        ++pos;
      }
 
      // '#'
      if (fmt[pos] == '#')
      {
        ++pos;
      }
 
      // '0'
      if (fmt[pos] == '0')
      {
        ++pos;
      }
 
      // width: number or nested replacement
      if (ct_is_digit(fmt[pos]))
      {
        if (ct_parse_num(fmt, pos) == -2)
          return false; // overflow
      }
      else if (fmt[pos] == '{')
      {
        if (!ct_parse_nested_replacement(fmt, pos, n_args, auto_count, has_auto, has_manual))
        {
          return false;
        }
      }
 
      // precision: '.' followed by number or nested replacement
      bool has_precision = false;
      if (fmt[pos] == '.')
      {
        has_precision = true;
        ++pos;
        if (ct_is_digit(fmt[pos]))
        {
          if (ct_parse_num(fmt, pos) == -2)
            return false; // overflow
        }
        else if (fmt[pos] == '{')
        {
          if (!ct_parse_nested_replacement(fmt, pos, n_args, auto_count, has_auto, has_manual))
          {
            return false;
          }
        }
        // else: '.' with no precision number/replacement — still valid (empty precision)
      }
 
      // locale-specific: 'L'
      if (fmt[pos] == 'L')
      {
        ++pos;
      }
 
      // type
      if (ct_is_type(fmt[pos]))
      {
        parsed_type = fmt[pos];
        ++pos;
      }
 
      // After parsing the spec, we must be at '}' (the closing brace is consumed by the caller)
      // Any remaining characters before '}' means invalid spec
      if (fmt[pos] != '}')
      {
        return false;
      }
 
      parsed_has_precision = has_precision;
 
      return true;
    }
  } // namespace private_format_check
 
  template <class... Args>
  constexpr bool check_format(const char* fmt)
  {
    const int n_args     = static_cast<int>(sizeof...(Args));
    int       pos        = 0;
    int       auto_count = 0;
    bool      has_auto   = false;
    bool      has_manual = false;
 
    // Build a constexpr array mapping arg index -> type category
    const private_format_check::type_category arg_categories[] = {
      private_format_check::get_type_category<Args>()...,
      private_format_check::type_category::NONE // sentinel for zero-arg case
    };
 
    while (fmt[pos] != '\0')
    {
      char c = fmt[pos];
      ++pos;
 
      if (c == '{')
      {
        if (fmt[pos] == '{')
        {
          // escaped '{'
          ++pos;
          continue;
        }
 
        // Start of a replacement field: [arg_id][:format_spec]
        int resolved_index = -1;
        int arg_index      = private_format_check::ct_parse_num(fmt, pos);
        if (arg_index == -2)
          return false; // overflow in arg index
        if (arg_index >= 0)
        {
          // manual index
          if (has_auto)
            return false; // mixing auto and manual
          has_manual = true;
          if (arg_index >= n_args)
            return false; // index out of range
          resolved_index = arg_index;
        }
        else
        {
          // automatic index
          if (has_manual)
            return false; // mixing auto and manual
          has_auto = true;
          if (auto_count >= n_args)
            return false; // too many arguments
          resolved_index = auto_count;
          ++auto_count;
        }
 
        char type_char     = '\0';
        bool has_precision = false;
        if (fmt[pos] == ':')
        {
          ++pos; // skip ':'
          if (!private_format_check::ct_skip_format_spec(fmt, pos, n_args, auto_count, has_auto, has_manual, type_char, has_precision))
          {
            return false;
          }
        }
 
        // Validate type specifier against argument type
        if (resolved_index >= 0 && resolved_index < n_args)
        {
          if (!private_format_check::ct_check_type_spec(arg_categories[resolved_index], type_char))
          {
            return false;
          }
 
          // Precision is not allowed for integer, boolean, or pointer types
          if (has_precision)
          {
            auto cat = arg_categories[resolved_index];
            if (cat == private_format_check::type_category::INTEGER || cat == private_format_check::type_category::BOOLEAN
                || cat == private_format_check::type_category::POINTER)
            {
              return false;
            }
            // Precision is also invalid for char when presented as char (not as integer)
            if (cat == private_format_check::type_category::CHAR && (type_char == '\0' || type_char == 'c' || type_char == '?'))
            {
              return false;
            }
          }
        }
 
        if (fmt[pos] != '}')
        {
          return false; // missing closing brace
        }
        ++pos; // skip '}'
      }
      else if (c == '}')
      {
        if (fmt[pos] != '}')
        {
          return false; // unmatched '}'
        }
        ++pos; // skip second '}'
      }
    }
 
    return true;
  }
 
  inline void please_note_this_is_error_message_format_string_syntax_error() noexcept {}
  #endif // ETL_USING_CPP20
 
  template <class... Args>
  struct basic_format_string
  {
    inline ETL_CONSTEVAL basic_format_string(const char* fmt)
      : _sv(fmt)
    {
  #if ETL_USING_CPP20
      // Compile-time validation: check_format runs at compile time via consteval.
      // In pre-C++20, runtime checks in vformat_to/parse_format_spec/etc. are sufficient.
      if (!check_format<Args...>(fmt))
      {
        // Calling a non-constexpr function in a consteval context triggers a compile error.
        please_note_this_is_error_message_format_string_syntax_error();
      }
  #endif
    }
 
    ETL_CONSTEXPR basic_format_string(const basic_format_string& other) = default;
    ETL_CONSTEXPR14 basic_format_string& operator=(const basic_format_string& other) = default;
 
    ETL_CONSTEXPR string_view get() const
    {
      return _sv;
    }
 
  private:
 
    string_view _sv;
  };
 
  template <class... Args>
  using format_string = basic_format_string<type_identity_t<Args>...>;
 
  template <class CharT>
  class basic_format_parse_context
  {
  public:
 
    using iterator       = string_view::const_iterator;
    using const_iterator = string_view::const_iterator;
    using char_type      = CharT;
 
    basic_format_parse_context(etl::string_view fmt, size_t n_args = 0)
      : range(fmt)
      , num_args(n_args)
      , current(0)
      , automatic_mode(false)
      , manual_mode(false)
    {
    }
 
    basic_format_parse_context<CharT>& operator=(const basic_format_parse_context&) = delete;
 
    iterator begin() const noexcept
    {
      return range.begin();
    }
 
    iterator end() const noexcept
    {
      return range.end();
    }
 
    ETL_CONSTEXPR14 void advance_to(iterator pos)
    {
      range = etl::string_view(pos, range.end());
    }
 
    ETL_CONSTEXPR14 size_t next_arg_id()
    {
      // automatic number generation only allowed if not already in manual mode
      ETL_ASSERT(manual_mode == false, ETL_ERROR(bad_format_string_exception));
      automatic_mode = true;
      ETL_ASSERT(current < num_args, ETL_ERROR(bad_format_string_exception) /* not enough arguments for generated index */);
      return current++;
    }
 
    ETL_CONSTEXPR14 void check_arg_id(size_t id)
    {
      // manual index specification only allowed if not already in automatic
      // mode
      ETL_ASSERT(automatic_mode == false, ETL_ERROR(bad_format_string_exception));
      manual_mode = true;
      ETL_ASSERT(id < num_args, ETL_ERROR(bad_format_string_exception) /* index out of range */);
    }
 
  private:
 
    etl::string_view range;
    size_t           num_args;
    size_t           current;
    bool             automatic_mode;
    bool             manual_mode;
 
    template <class, class>
    friend struct formatter;
  };
 
  using format_parse_context = basic_format_parse_context<char>;
 
  // Forward declaration of the formatter primary template (defined later). This
  // lets basic_format_arg detect user-provided formatter specialisations for
  // custom types and route them through basic_format_arg::handle.
  template <class T, class CharT = char>
  struct formatter;
 
  // Forward declaration of the format context (defined later) so that
  // is_formattable can probe for a usable formatter<T>::format() member, which
  // is templated on the output iterator type.
  template <class OutputIt, class CharT>
  class basic_format_context;
 
  namespace private_format
  {
    // Detects whether etl::formatter<T> exposes a usable parse() member.
    template <typename T, typename = void>
    struct has_formatter_parse : etl::false_type
    {
    };
 
    template <typename T>
    struct has_formatter_parse<T, etl::void_t<decltype(etl::declval<etl::formatter<T>&>().parse(etl::declval<format_parse_context&>()))> >
      : etl::true_type
    {
    };
 
    // Detects whether etl::formatter<T> exposes a usable format() member.
    // format() is templated on the output iterator, so it is probed with a
    // representative char* output iterator context.
    template <typename T, typename = void>
    struct has_formatter_format : etl::false_type
    {
    };
 
    template <typename T>
    struct has_formatter_format<T, etl::void_t<decltype(etl::declval<etl::formatter<T>&>().format(
                                     etl::declval<const T&>(), etl::declval<etl::basic_format_context<char*, char>&>()))> > : etl::true_type
    {
    };
 
    // The primary formatter template is empty, so a type is only formattable
    // when a (user-)provided specialisation supplies both a usable parse() and
    // a usable format() member - mirroring the std::formattable requirements.
    template <typename T>
    struct is_formattable : etl::bool_constant<has_formatter_parse<T>::value && has_formatter_format<T>::value>
    {
    };
  } // namespace private_format
 
  template <class Context>
  class basic_format_arg
  {
  public:
 
    // Type-erased wrapper that allows user-defined types to be formatted via an
    // etl::formatter<T> specialisation, mirroring std::basic_format_arg::handle.
    class handle
    {
    public:
 
      template <typename T>
      explicit handle(const T& value)
        : obj(static_cast<const void*>(etl::addressof(value)))
        , func(&format_custom_type<T>)
      {
      }
 
      void format(etl::basic_format_parse_context<char>& parse_ctx, Context& format_ctx) const
      {
        func(parse_ctx, format_ctx, obj);
      }
 
    private:
 
      template <typename T>
      static void format_custom_type(etl::basic_format_parse_context<char>& parse_ctx, Context& format_ctx, const void* ptr)
      {
        const T&          value = *static_cast<const T*>(ptr);
        etl::formatter<T> f;
        parse_ctx.advance_to(f.parse(parse_ctx));
        format_ctx.advance_to(f.format(value, format_ctx));
      }
 
      const void* obj;
      typedef void (*function_type)(etl::basic_format_parse_context<char>&, Context&, const void*);
      function_type func;
    };
 
    basic_format_arg() {}
 
    basic_format_arg(const bool v)
      : data(v)
    {
    }
 
    basic_format_arg(const int v)
      : data(v)
    {
    }
 
    basic_format_arg(const short v)
      : data(static_cast<int>(v))
    {
    }
 
    basic_format_arg(const unsigned short v)
      : data(static_cast<unsigned int>(v))
    {
    }
 
    basic_format_arg(const long int v)
      : data(static_cast<long long int>(v))
    {
    }
 
    basic_format_arg(const unsigned int v)
      : data(v)
    {
    }
 
    basic_format_arg(const long long int v)
      : data(v)
    {
    }
 
    basic_format_arg(const unsigned long long int v)
      : data(v)
    {
    }
 
    // Additional type to list of basic types as defined for
    // std::basic_format_arg: Mapping unsigned long to unsigned long long int
    basic_format_arg(const unsigned long v)
      : data(static_cast<unsigned long long int>(v))
    {
    }
 
    basic_format_arg(const char* v)
      : data(v)
    {
    }
 
    basic_format_arg(char v)
      : data(v)
    {
    }
 
    basic_format_arg(const signed char v)
      : data(static_cast<char>(v))
    {
    }
 
    basic_format_arg(const unsigned char v)
      : data(static_cast<char>(v))
    {
    }
 
  #if ETL_USING_FORMAT_FLOATING_POINT
    basic_format_arg(const float v)
      : data(v)
    {
    }
 
    basic_format_arg(const double v)
      : data(v)
    {
    }
 
    basic_format_arg(const long double v)
      : data(v)
    {
    }
  #endif
 
    basic_format_arg(const etl::string_view v)
      : data(v)
    {
    }
 
    basic_format_arg(const etl::ibasic_string<char>& v)
      : data(etl::string_view(v.data(), v.size()))
    {
    }
 
    basic_format_arg(const basic_format_arg& other)
      : data(other.data)
    {
    }
 
    basic_format_arg(const void* v)
      : data(v)
    {
    }
 
    // Converting constructor for user-defined types that provide an
    // etl::formatter<T> specialisation. The value is stored type-erased in a
    // handle, matching the std::basic_format_arg behaviour for custom types.
    template <typename T, typename = etl::enable_if_t<private_format::is_formattable<T>::value> >
    basic_format_arg(const T& v)
      : data(handle(v))
    {
    }
 
    basic_format_arg& operator=(const basic_format_arg& other)
    {
      data = other.data;
      return *this;
    }
 
    explicit operator bool() const
    {
      return !etl::holds_alternative<etl::monostate>(data);
    }
 
    template <class R, class Visitor>
    R visit(Visitor&& vis)
    {
      return etl::visit(etl::forward<Visitor>(vis), data);
    }
 
  private:
 
    // Storage for the argument value.
    //
    // This is the limited number of types as defined in std::basic_format_arg
    // https://en.cppreference.com/w/cpp/utility/format/basic_format_arg.html
    //
    // Further types to be supported are mapped onto these via the converting
    // constructors above. User-defined types are stored type-erased in the
    // special handle member, which formats them through their etl::formatter
    // specialisation.
    etl::variant< etl::monostate, bool, char, int, unsigned int, long long int, unsigned long long int,
  #if ETL_USING_FORMAT_FLOATING_POINT
                  float, double, long double,
  #endif
                  const char*, etl::string_view, const void*, handle >
      data;
  };
 
  template <class Context, class... Args>
  class format_arg_store
  {
  public:
 
    format_arg_store(Args&... args)
      : _args{args...}
    {
    }
 
    basic_format_arg<Context> get(size_t i) const
    {
      return _args.get(i);
    }
 
    etl::array_view<basic_format_arg<Context>> get()
    {
      return _args;
    }
 
  private:
 
    etl::array<basic_format_arg<Context>, sizeof...(Args)> _args;
  };
 
  template <class Context>
  class basic_format_args
  {
  public:
 
    template <class... Args>
    basic_format_args(format_arg_store<Context, Args...>& store)
      : _args(store.get())
    {
    }
 
    basic_format_args(const basic_format_args<Context>& other)
      : _args(other._args)
    {
    }
 
    basic_format_args& operator=(const basic_format_args<Context>& other)
    {
      _args = other._args;
      return *this;
    }
 
    basic_format_arg<Context> get(size_t i) const
    {
      return _args[i];
    }
 
    // non-standard
    size_t size()
    {
      return _args.size();
    }
 
  private:
 
    etl::array_view<basic_format_arg<Context>> _args;
  };
 
  namespace private_format
  {
    using char_type = char;
 
    enum class spec_align_t
    {
      NONE, // default
      START,
      END,
      CENTER
    };
 
    enum class spec_sign_t
    {
      MINUS, // default
      PLUS,
      SPACE
    };
 
    struct format_spec_t
    {
      etl::optional<size_t> index{etl::nullopt_t()};
      spec_align_t          align{spec_align_t::NONE};         // '<' / '>' / '^' / none (default)
      char_type             fill{' '};                         // fill character (' ' is default)
      spec_sign_t           sign{spec_sign_t::MINUS};          // '+' / '-' (default) / ' '
      bool                  hash{false};                       // #
      bool                  zero{false};                       // 0
      etl::optional<size_t> width{etl::nullopt_t()};           // the arg index if width_nested_replacement == true
      bool                  width_nested_replacement{false};   // {}
      etl::optional<size_t> precision{etl::nullopt_t()};       // the arg index if
                                                               // precision_nested_replacement == true
      bool                precision_nested_replacement{false}; // {}
      bool                locale_specific{false};              // 'L'
      etl::optional<char> type{etl::nullopt_t()};              // literal 's', 'b', 'd', ...
    };
  } // namespace private_format
 
  template <class OutputIt, class CharT>
  class basic_format_context
  {
  public:
 
    using iterator  = OutputIt;
    using char_type = CharT;
 
    basic_format_context(const basic_format_context& other)
      : _it(other._it)
      , _format_args(other._format_args)
    {
    }
 
    basic_format_context(OutputIt it, basic_format_args<basic_format_context>& fmt_args)
      : _it(it)
      , _format_args(fmt_args)
    {
    }
 
    basic_format_context& operator=(const basic_format_context&) = delete;
 
    basic_format_arg<basic_format_context> arg(size_t id) const
    {
      return _format_args.get(id);
    }
 
    iterator out()
    {
      return _it;
    }
 
    void advance_to(iterator it)
    {
      _it = it;
    }
 
    private_format::format_spec_t format_spec;
 
  private:
 
    iterator                                 _it;
    basic_format_args<basic_format_context>& _format_args;
  };
 
  template <class OutputIt>
  using format_context = basic_format_context<OutputIt, char>;
 
  template <class OutputIt>
  using format_args = basic_format_args<format_context<OutputIt>>;
 
  template <class OutputIt>
  using format_arg = basic_format_arg<format_context<OutputIt>>;
 
  template <class OutputIt, class Context = format_context<OutputIt>, class... Args>
  format_arg_store<Context, Args...> make_format_args(Args&... args)
  {
    return format_arg_store<Context, Args...>(args...);
  }
 
  namespace private_format
  {
    inline bool is_digit(const char c)
    {
      return c >= '0' && c <= '9';
    }
 
    inline void advance(format_parse_context& parse_ctx)
    {
      parse_ctx.advance_to(parse_ctx.begin() + 1);
    }
 
    inline etl::optional<size_t> parse_num(format_parse_context& parse_ctx)
    {
      etl::optional<size_t> result;
      auto                  fmt_it = parse_ctx.begin();
      while (fmt_it != parse_ctx.end())
      {
        const char c = *fmt_it;
        if (is_digit(c))
        {
          size_t old_value = result.value_or(0);
          size_t new_value = old_value * 10 + static_cast<size_t>(c - '0');
          if (new_value < old_value)
          {
            // Overflow detected
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
          }
          result = new_value;
        }
        else
        {
          break;
        }
        ++fmt_it;
      }
      if (result.has_value())
      {
        parse_ctx.advance_to(fmt_it);
      }
      return result;
    }
 
    inline etl::optional<char> parse_any_of(format_parse_context& parse_ctx, etl::string_view chars)
    {
      etl::optional<char> result;
      auto                fmt_it = parse_ctx.begin();
      if (fmt_it != parse_ctx.end())
      {
        const char c  = *fmt_it;
        auto       it = etl::find(chars.cbegin(), chars.cend(), c);
        if (it != chars.cend())
        {
          result = *it;
          ++fmt_it;
          parse_ctx.advance_to(fmt_it);
        }
      }
      return result;
    }
 
    inline bool parse_char(format_parse_context& parse_ctx, char c)
    {
      auto fmt_it = parse_ctx.begin();
      if (fmt_it != parse_ctx.end())
      {
        char value = *fmt_it;
        if (value == c)
        {
          ++fmt_it;
          parse_ctx.advance_to(fmt_it);
          return true;
        }
      }
      return false;
    }
 
    inline bool is_align_character(char c)
    {
      return c == '<' || c == '>' || c == '^';
    }
 
    inline spec_align_t align_from_char(char c)
    {
      spec_align_t result = spec_align_t::NONE;
      switch (c)
      {
        case '<': result = spec_align_t::START; break;
        case '>': result = spec_align_t::END; break;
        case '^': result = spec_align_t::CENTER; break;
        default:
          // invalid alignment specification
          ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
      }
      return result;
    }
 
    inline spec_align_t parse_fill_and_align(format_parse_context& parse_ctx, char_type& fill)
    {
      spec_align_t result = spec_align_t::NONE;
      fill                = ' '; // default
 
      auto fmt_it = parse_ctx.begin();
      if (fmt_it != parse_ctx.end())
      {
        const char c = *fmt_it;
        ++fmt_it;
 
        if (is_align_character(c))
        {
          result = align_from_char(c);
          parse_ctx.advance_to(fmt_it);
        }
        else if (fmt_it != parse_ctx.end())
        {
          const char c2 = *fmt_it;
          ++fmt_it;
          if (is_align_character(c2))
          {
            result = align_from_char(c2);
            ETL_ASSERT(c != '{' && c != '}',
                       ETL_ERROR(bad_format_string_exception)); // no { or } allowed as
                                                                // fill character
            fill = c;
            parse_ctx.advance_to(fmt_it);
          }
        }
        else
        {
          // no align and fill spec (valid)
        }
      }
      return result;
    }
 
    inline spec_sign_t sign_from_char(const char c)
    {
      spec_sign_t result = spec_sign_t::MINUS;
      switch (c)
      {
        case '-': result = spec_sign_t::MINUS; break;
        case '+': result = spec_sign_t::PLUS; break;
        case ' ': result = spec_sign_t::SPACE; break;
        default:
          // invalid sign character c
          ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
      }
      return result;
    }
 
    inline bool parse_nested_replacement(format_parse_context& parse_ctx, etl::optional<size_t>& index)
    {
      bool start = parse_char(parse_ctx, '{');
      if (start)
      {
        auto num = parse_num(parse_ctx);
        if (num)
        {
          // manual mode
          index = num;
          parse_ctx.check_arg_id(*index);
          bool end = parse_char(parse_ctx, '}');
          if (end)
          {
            return true;
          }
          else
          {
            // bad nested replacement index spec
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
          }
        }
        else
        {
          bool end = parse_char(parse_ctx, '}');
          if (end)
          {
            index = parse_ctx.next_arg_id();
            return true;
          }
          else
          {
            // bad nested replacement spec
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
          }
        }
      }
      return false;
    }
 
    template <class OutputIt>
    void parse_format_spec(format_parse_context& parse_ctx, format_context<OutputIt>& fmt_context)
    {
      auto& format_spec = fmt_context.format_spec;
 
      format_spec = format_spec_t(); // reset format_spec to defaults
 
      format_spec.index = parse_num(parse_ctx); // optional explicit index
 
      // Consume the value's auto-index before parsing the format spec body,
      // so that nested replacement fields for width/precision get correct
      // auto-indices. Per C++ standard, in {:{}}, the value arg is consumed
      // first (arg 0), then the width arg (arg 1).
      if (!format_spec.index.has_value())
      {
        format_spec.index = parse_ctx.next_arg_id();
      }
      else
      {
        parse_ctx.check_arg_id(*format_spec.index);
      }
 
      bool colon = parse_char(parse_ctx, ':');
      if (colon)
      {
        format_spec.align = parse_fill_and_align(parse_ctx, format_spec.fill);
 
        etl::optional<char> sign = parse_any_of(parse_ctx, "+- ");
        if (sign)
        {
          format_spec.sign = sign_from_char(*sign);
        }
 
        format_spec.hash = parse_char(parse_ctx, '#');
        format_spec.zero = parse_char(parse_ctx, '0');
 
        format_spec.width = parse_num(parse_ctx);
        if (!format_spec.width)
        {
          // possibly with via nested replacement
          format_spec.width_nested_replacement = parse_nested_replacement(parse_ctx, format_spec.width);
        }
 
        if (parse_char(parse_ctx, '.'))
        {
          format_spec.precision = parse_num(parse_ctx);
          if (!format_spec.precision)
          {
            // possibly with via nested replacement
            format_spec.precision_nested_replacement = parse_nested_replacement(parse_ctx, format_spec.precision);
          }
        }
 
        format_spec.locale_specific = parse_char(parse_ctx, 'L');
 
        format_spec.type = parse_any_of(parse_ctx, "s?bBcdoxXaAeEfFgGpP");
      }
    }
  } // namespace private_format
 
  template <class T, class CharT>
  struct formatter
  {
    using char_type = CharT;
  };
 
  template <>
  struct formatter<etl::monostate>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      return parse_ctx.end();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(etl::monostate arg, format_context<OutputIt>& fmt_ctx)
    {
      (void)arg;
      return fmt_ctx.out();
    }
  };
 
  namespace private_format
  {
    // for 4321, return 1000
    template <typename UnsignedT, typename = etl::enable_if_t<etl::is_unsigned<UnsignedT>::value>>
    UnsignedT get_highest_digit(UnsignedT value, size_t base = 10)
    {
      ETL_ASSERT(base > 1, ETL_ERROR(bad_format_string_exception));
      UnsignedT result = 1;
      value /= base;
      while (result <= value)
      {
        result *= base;
      }
      return result;
    }
 
    template <typename T>
    T int_pow(T base, T exp)
    {
      T result = 1;
      while (exp > 0)
      {
        if (exp % 2 == 1)
          result *= base;
        base *= base;
        exp /= 2;
      }
      return result;
    }
 
    template <typename OutputIt, typename T>
    void format_sign(OutputIt& it, T value, const format_spec_t& spec)
    {
      char c = '\0';
      if (value < 0)
      {
        c = '-';
      }
      else
      {
        switch (spec.sign)
        {
          case spec_sign_t::MINUS:
            // c already set above if negative
            break;
          case spec_sign_t::PLUS: c = '+'; break;
          case spec_sign_t::SPACE: c = ' '; break;
          default:
            // invalid sign
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
 
      if (c != '\0')
      {
        *it = c;
        ++it;
      }
    }
 
    template <typename OutputIt>
    void format_sequence(OutputIt& out_it, etl::string_view value)
    {
      auto it = value.cbegin();
      while (it != value.cend())
      {
        *out_it = *it;
        ++it;
        ++out_it;
      }
    }
 
    template <typename OutputIt, typename T>
    void format_alternate_form(OutputIt& it, T value, const format_spec_t& spec)
    {
      if (spec.hash && spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 'b': format_sequence(it, "0b"); break;
          case 'B': format_sequence(it, "0B"); break;
          case 'o':
            // Per C++ standard, # for octal adds leading 0 only if not already present
            if (value != 0)
            {
              format_sequence(it, "0");
            }
            break;
          case 'x': format_sequence(it, "0x"); break;
          case 'X':
            format_sequence(it, "0X");
            break;
            // default: no prefix
        }
      }
    }
 
    template <typename OutputIt>
    void format_plain_char(OutputIt& it, char_type c)
    {
      *it = c;
      ++it;
    }
 
    template <typename OutputIt>
    void format_escaped_char(OutputIt& it, char_type c)
    {
      switch (c)
      {
        case '\t': format_sequence(it, "\\t"); break;
        case '\n': format_sequence(it, "\\n"); break;
        case '\r': format_sequence(it, "\\r"); break;
        case '"': format_sequence(it, "\\\""); break;
        case '\'': format_sequence(it, "\\'"); break;
        case '\\': format_sequence(it, "\\\\"); break;
        default: *it = c; ++it;
      }
    }
 
    template <typename OutputIt>
    void fill(OutputIt& it, size_t size, char_type c)
    {
      while (size > 0)
      {
        *it = c;
        ++it;
        --size;
      }
    }
 
    template <size_t default_base = 10>
    inline size_t base_from_spec(const format_spec_t& spec)
    {
      size_t base = default_base;
      if (spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 'a':
          case 'A': base = 16; break;
          case 'b':
          case 'B': base = 2; break;
          case 'o': base = 8; break;
          case 'p':
          case 'P':
          case 'x':
          case 'X':
            base = 16;
            break;
            // default: no prefix
        }
      }
      return base;
    }
 
    inline bool is_uppercase(const char c)
    {
      return c >= 'A' && c <= 'Z';
    }
 
    template <typename OutputIt, typename T>
    void format_digit_char(OutputIt& it, T value, const format_spec_t& spec)
    {
      if (value <= 9)
      {
        *it = static_cast<char_type>('0' + static_cast<typename etl::make_unsigned<T>::type>(value));
      }
      else
      {
        if (spec.type.has_value() && is_uppercase(spec.type.value()))
        {
          *it = static_cast<char_type>('A' + static_cast<typename etl::make_unsigned<T>::type>(value - 10));
        }
        else
        {
          *it = static_cast<char_type>('a' + static_cast<typename etl::make_unsigned<T>::type>(value - 10));
        }
      }
      ++it;
    }
 
    inline void adjust_width_from_spec(const format_spec_t& spec, size_t& width)
    {
      if (spec.zero && spec.width.has_value())
      {
        width = etl::max(width, spec.width.value());
      }
    }
 
    inline void check_precision(const format_spec_t& spec)
    {
      if (spec.precision.has_value())
      {
        // precision not allowed for integer numbers
        ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
      }
    }
 
    // used for both integers and float parts
    // skip_last_zeros helps in case of printing after-the-decimal zeros which
    // are redundant then
    template <typename OutputIt, typename T, T default_base = 10, bool skip_last_zeros = false>
    void format_plain_num(OutputIt& it, T value, const format_spec_t& spec, size_t width = 0)
    {
      using UnsignedT = typename etl::make_unsigned<T>::type;
 
      UnsignedT unsigned_value = etl::absolute_unsigned(value);
 
      size_t    base          = base_from_spec<default_base>(spec);
      UnsignedT highest_digit = get_highest_digit<UnsignedT>(unsigned_value, base);
      if (width > 0)
      {
        UnsignedT align_highest_digit = int_pow<UnsignedT>(base, width - 1);
        highest_digit                 = etl::max<UnsignedT>(align_highest_digit, highest_digit);
      }
 
      // this loop is iterated at least once, to print a number
      while (highest_digit > 0)
      {
        UnsignedT digit = unsigned_value / highest_digit;
        unsigned_value %= highest_digit;
        format_digit_char(it, digit, spec);
 
        if ETL_IF_CONSTEXPR (skip_last_zeros)
        {
          if (unsigned_value == 0)
          {
            break;
          }
        }
 
        highest_digit /= base;
      }
    }
 
    // for integers
    template <typename OutputIt, typename T, bool skip_last_zeros = false>
    void format_num(OutputIt& it, T value, const format_spec_t& spec)
    {
      size_t width = 0;
      format_sign<OutputIt, T>(it, value, spec);
      format_alternate_form<OutputIt, T>(it, value, spec);
      adjust_width_from_spec(spec, width);
      check_precision(spec);
      format_plain_num(it, value, spec, width);
    }
 
  #if ETL_USING_FORMAT_FLOATING_POINT
    #if ETL_NOT_USING_FORMAT_LONG_DOUBLE_MATH
    //***********************************
    // Math function wrappers to handle toolchains that don't provide
    // long double math functions (log10l, floorl, powl, modfl, roundl).
    // When ETL_FORMAT_NO_LONG_DOUBLE_MATH is defined, long double overloads
    // cast through double. For float and double, the standard functions are
    // called directly via the template versions.
    //***********************************
    inline long double format_log10(long double value)
    {
      return static_cast<long double>(::log10(static_cast<double>(value)));
    }
    inline long double format_floor(long double value)
    {
      return static_cast<long double>(::floor(static_cast<double>(value)));
    }
    inline long double format_pow(long double base, long double exp)
    {
      return static_cast<long double>(::pow(static_cast<double>(base), static_cast<double>(exp)));
    }
    inline long double format_round(long double value)
    {
      return static_cast<long double>(::round(static_cast<double>(value)));
    }
    inline long double format_modf(long double value, long double* iptr)
    {
      double d_iptr;
      double result = ::modf(static_cast<double>(value), &d_iptr);
      *iptr         = static_cast<long double>(d_iptr);
      return static_cast<long double>(result);
    }
    #endif
 
    template <typename T>
    T format_log10(T value)
    {
      return ::log10(value);
    }
    template <typename T>
    T format_floor(T value)
    {
      return ::floor(value);
    }
    template <typename T>
    T format_pow(T base, T exp)
    {
      return ::pow(base, exp);
    }
    template <typename T>
    T format_round(T value)
    {
      return ::round(value);
    }
    template <typename T>
    T format_modf(T value, T* iptr)
    {
      return ::modf(value, iptr);
    }
 
    template <typename OutputIt, typename T>
    void format_floating_default(OutputIt& it, T value, const format_spec_t& spec)
    {
      const size_t fractional_decimals = 6; // default
 
      // Detect sign using signbit to correctly handle -0.0
      bool sign      = signbit(value);
      T    abs_value = sign ? -value : value;
 
      // Use scientific notation for values that would overflow unsigned long long
      // or that are too small for meaningful fixed-point digits
      if (abs_value >= static_cast<T>(1e18) || (abs_value > static_cast<T>(0) && abs_value < static_cast<T>(1e-6)))
      {
        format_spec_t spec_e = spec;
        spec_e.type          = 'e';
        format_floating_e(it, value, spec_e);
        return;
      }
 
      T integral;
      T fractional = format_modf(value, &integral);
 
      // Take absolute values to avoid casting negative values to unsigned
      if (sign)
      {
        fractional = -fractional;
        integral   = -integral;
      }
 
      unsigned long long int scale          = int_pow<unsigned long long int>(10, fractional_decimals);
      unsigned long long int fractional_int = static_cast<unsigned long long int>(format_round(fractional * scale));
      unsigned long long int integral_int   = static_cast<unsigned long long int>(integral);
 
      if (fractional_int == scale)
      {
        fractional_int = 0;
        ++integral_int;
      }
 
      private_format::format_sign<OutputIt, int>(it, sign ? -1 : 0, spec);
      private_format::format_plain_num<OutputIt, unsigned long long int>(it, integral_int, spec);
      private_format::format_sequence<OutputIt>(it, ".");
      private_format::format_plain_num<OutputIt, unsigned long long int, 10, true>(it, fractional_int, spec, fractional_decimals);
    }
 
    // floating point in hex notation
    template <typename OutputIt, typename T>
    void format_floating_a(OutputIt& it, T value, const format_spec_t& spec)
    {
      static const size_t fractional_decimals = 10; // default
      static const size_t exponent_decimals   = 1;
      long long int       exponent_int        = 0;
 
      // Detect sign using signbit to correctly handle -0.0
      bool sign = signbit(value);
 
      T integral;
      T fractional = format_modf(value, &integral);
 
      while (value >= 0x10 || value <= -0x10)
      {
        ++exponent_int;
        value /= 0x10;
        fractional = format_modf(value, &integral);
      }
 
      while ((value > 0.0000000000001 && value < 1) || (value < -0.0000000000001 && value > -1))
      {
        --exponent_int;
        value *= 0x10;
        fractional = format_modf(value, &integral);
      }
 
      // Take absolute values to avoid casting negative values to unsigned
      if (sign)
      {
        fractional = -fractional;
        integral   = -integral;
      }
 
      unsigned long long int scale          = int_pow<unsigned long long int>(0x10, fractional_decimals);
      unsigned long long int fractional_int = static_cast<unsigned long long int>(format_round(fractional * scale));
      unsigned long long int integral_int   = static_cast<unsigned long long int>(integral);
 
      if (fractional_int == scale)
      {
        fractional_int = 0;
        ++integral_int;
      }
 
      private_format::format_sign<OutputIt, int>(it, sign ? -1 : 0, spec);
      private_format::format_plain_char<OutputIt>(it, '0');
      char hex_letter = 'x';
      if (is_uppercase(spec.type.value()))
      {
        hex_letter = 'X';
      }
      private_format::format_plain_char<OutputIt>(it, hex_letter);
      private_format::format_plain_num<OutputIt, unsigned long long int, 16>(it, integral_int, spec);
      private_format::format_plain_char<OutputIt>(it, '.');
      private_format::format_plain_num<OutputIt, unsigned long long int, 16, true>(it, fractional_int, spec, fractional_decimals);
      char letter = 'p';
      if (is_uppercase(spec.type.value()))
      {
        letter = 'P';
      }
      private_format::format_plain_char<OutputIt>(it, letter);
      private_format::format_plain_char<OutputIt>(it, (exponent_int < 0) ? '-' : '+');
      private_format::format_plain_num<OutputIt, long long int, 16>(it, exponent_int, spec, exponent_decimals);
    }
 
    template <typename OutputIt, typename T>
    void format_floating_e(OutputIt& it, T value, const format_spec_t& spec)
    {
      static const size_t fractional_decimals = 6; // default
      static const size_t exponent_decimals   = 2;
      long long int       exponent_int        = 0;
 
      // Detect sign using signbit to correctly handle -0.0
      bool sign = signbit(value);
 
      T abs_value = sign ? -value : value;
 
      if (abs_value > static_cast<T>(0))
      {
        exponent_int = static_cast<long long int>(format_floor(format_log10(abs_value)));
        value        = abs_value / format_pow(static_cast<T>(10), static_cast<T>(exponent_int));
        // Correct for floating-point rounding in log10/pow
        if (value >= static_cast<T>(10))
        {
          value /= static_cast<T>(10);
          ++exponent_int;
        }
        else if (value < static_cast<T>(1))
        {
          value *= static_cast<T>(10);
          --exponent_int;
        }
      }
      else
      {
        value = static_cast<T>(0);
      }
 
      T integral;
      T fractional = format_modf(value, &integral);
 
      unsigned long long int scale          = int_pow<unsigned long long int>(10, fractional_decimals);
      unsigned long long int fractional_int = static_cast<unsigned long long int>(format_round(fractional * scale));
      unsigned long long int integral_int   = static_cast<unsigned long long int>(integral);
 
      if (fractional_int == scale)
      {
        fractional_int = 0;
        ++integral_int;
 
        if (integral_int == 10)
        {
          integral_int = 1;
          ++exponent_int;
        }
      }
 
      private_format::format_sign<OutputIt, int>(it, sign ? -1 : 0, spec);
      private_format::format_plain_num<OutputIt, unsigned long long int>(it, integral_int, spec);
      private_format::format_sequence<OutputIt>(it, ".");
      private_format::format_plain_num<OutputIt, unsigned long long int>(it, fractional_int, spec, fractional_decimals);
      char letter = 'e';
      if (is_uppercase(spec.type.value()))
      {
        letter = 'E';
      }
      private_format::format_plain_char<OutputIt>(it, letter);
      private_format::format_plain_char<OutputIt>(it, (exponent_int < 0) ? '-' : '+');
      private_format::format_plain_num<OutputIt, long long int>(it, exponent_int, spec, exponent_decimals);
    }
 
    template <typename OutputIt, typename T>
    void format_floating_f(OutputIt& it, T value, const format_spec_t& spec)
    {
      const size_t fractional_decimals = 6; // default
 
      // Detect sign using signbit to correctly handle -0.0
      bool sign = signbit(value);
 
      T integral;
      T fractional = format_modf(value, &integral);
 
      // Take absolute values to avoid casting negative values to unsigned
      if (sign)
      {
        fractional = -fractional;
        integral   = -integral;
      }
 
      unsigned long long int scale          = int_pow<unsigned long long int>(10, fractional_decimals);
      unsigned long long int fractional_int = static_cast<unsigned long long int>(format_round(fractional * scale));
      unsigned long long int integral_int   = static_cast<unsigned long long int>(integral);
 
      if (fractional_int == scale)
      {
        fractional_int = 0;
        ++integral_int;
      }
 
      private_format::format_sign<OutputIt, int>(it, sign ? -1 : 0, spec);
      private_format::format_plain_num<OutputIt, unsigned long long int>(it, integral_int, spec);
      private_format::format_sequence<OutputIt>(it, ".");
      private_format::format_plain_num<OutputIt, unsigned long long int>(it, fractional_int, spec, fractional_decimals);
    }
  #endif
 
    class dummy_assign_to
    {
    public:
 
      dummy_assign_to& operator=(char_type)
      {
        return *this;
      }
    };
 
    template <class OutputIt>
    class limit_assign_to
    {
    public:
 
      limit_assign_to(OutputIt o, bool is_active)
        : out(o)
        , active(is_active)
      {
      }
 
      limit_assign_to& operator=(char_type c)
      {
        if (active)
        {
          *out = c;
        }
        return *this;
      }
 
    private:
 
      OutputIt out;
      bool     active;
    };
 
    template <class OutputIt>
    class limit_iterator
    {
    public:
 
      limit_iterator(OutputIt& it, size_t n)
        : out(it)
        , limit(n)
      {
      }
 
      limit_iterator(const limit_iterator& other)            = default;
      limit_iterator(limit_iterator&& other)                 = default;
      limit_iterator& operator=(const limit_iterator& other) = default;
      limit_iterator& operator=(limit_iterator&& other)      = default;
 
      limit_assign_to<OutputIt> operator*()
      {
        return limit_assign_to<OutputIt>(out, (limit > 0));
      }
 
      limit_iterator& operator++()
      {
        if (limit > 0)
        {
          --limit;
          ++out;
        }
        return *this;
      }
 
      limit_iterator operator++(int)
      {
        limit_iterator temp = *this;
        if (limit > 0)
        {
          --limit;
          out++;
        }
        return temp;
      }
 
      OutputIt get()
      {
        return out;
      }
 
    private:
 
      OutputIt out;
      size_t   limit;
    };
 
    class counter_iterator
    {
    public:
 
      counter_iterator()
        : count(0)
      {
      }
 
      counter_iterator(const counter_iterator& other)            = default;
      counter_iterator& operator=(const counter_iterator& other) = default;
 
      dummy_assign_to operator*()
      {
        return dummy_assign_to();
      }
 
      counter_iterator& operator++()
      {
        ++count;
        return *this;
      }
 
      counter_iterator operator++(int)
      {
        counter_iterator temp = *this;
        count++;
        return temp;
      }
 
      size_t value()
      {
        return count;
      }
 
    private:
 
      size_t count;
    };
 
  #if ETL_USING_FORMAT_FLOATING_POINT
    template <typename OutputIt, typename T>
    void format_floating_g(OutputIt& it, T value, const format_spec_t& spec)
    {
      private_format::counter_iterator counter_e, counter_f;
 
      format_floating_e(counter_e, value, spec);
      format_floating_f(counter_f, value, spec);
 
      if (counter_e.value() < counter_f.value())
      {
        format_floating_e(it, value, spec);
      }
      else
      {
        format_floating_f(it, value, spec);
      }
    }
 
    template <typename OutputIt, typename T>
    void format_floating(OutputIt& it, T value, const format_spec_t& spec)
    {
      if (isnan(value))
      {
        if (spec.type.has_value() && (is_uppercase(spec.type.value())))
        {
          format_sequence(it, "NAN");
        }
        else
        {
          format_sequence(it, "nan");
        }
      }
      else if (isinf(value))
      {
        if (spec.type.has_value() && (is_uppercase(spec.type.value())))
        {
          format_sequence(it, "INF");
        }
        else
        {
          format_sequence(it, "inf");
        }
      }
      else if (!spec.type.has_value())
      {
        format_floating_default(it, value, spec);
      }
      else
      {
        switch (spec.type.value())
        {
          case 'a':
          case 'A': format_floating_a(it, value, spec); break;
          case 'e':
          case 'E': format_floating_e(it, value, spec); break;
          case 'f':
          case 'F': format_floating_f(it, value, spec); break;
          case 'g':
          case 'G': format_floating_g(it, value, spec); break;
          default:
            // unknown presentation type
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
    }
  #endif
 
    template <class OutputIt>
    struct format_visitor
    {
      using output_iterator = OutputIt;
 
      format_visitor(format_parse_context& parse_context, format_context<OutputIt>& f_ctx)
        : parse_ctx(parse_context)
        , fmt_ctx(f_ctx)
      {
      }
 
      // for all the built-in alternatives stored in basic_format_arg
      template <typename T>
      void operator()(T value)
      {
        formatter<T>                   f;
        format_parse_context::iterator it = f.parse(parse_ctx);
        parse_ctx.advance_to(it);
        OutputIt fit = f.format(value, fmt_ctx);
        fmt_ctx.advance_to(fit);
      }
 
      // for user-defined types routed through basic_format_arg::handle
      void operator()(typename basic_format_arg<format_context<OutputIt> >::handle h)
      {
        h.format(parse_ctx, fmt_ctx);
      }
 
      format_parse_context&     parse_ctx;
      format_context<OutputIt>& fmt_ctx;
    };
 
    template <class OutputIt>
    void output(format_context<OutputIt>& fmt_context, char c)
    {
      *fmt_context.out() = c;
      OutputIt tmp       = fmt_context.out();
      tmp++;
      fmt_context.advance_to(tmp);
    }
 
    // Visitor to extract an integer value as size_t from a format arg (for nested replacement fields)
    struct size_t_extractor
    {
      size_t value;
 
      size_t_extractor()
        : value(0)
      {
      }
 
      void operator()(int v)
      {
        value = static_cast<size_t>(v);
      }
      void operator()(unsigned int v)
      {
        value = static_cast<size_t>(v);
      }
      void operator()(long long int v)
      {
        value = static_cast<size_t>(v);
      }
      void operator()(unsigned long long int v)
      {
        value = static_cast<size_t>(v);
      }
 
      // All other types are invalid for width/precision - ignore
      template <typename T>
      void operator()(T)
      {
      }
    };
 
    // Resolve nested replacement fields for width and precision in the format spec.
    // When width_nested_replacement or precision_nested_replacement is true, the
    // width/precision value holds the arg index, which must be resolved to the actual value.
    template <class OutputIt>
    void resolve_nested_replacements(format_spec_t& spec, format_args<OutputIt>& args)
    {
      if (spec.width_nested_replacement && spec.width.has_value())
      {
        format_arg<OutputIt> width_arg = args.get(spec.width.value());
        size_t_extractor     ext;
        width_arg.template visit<void>(ext);
        spec.width                    = ext.value;
        spec.width_nested_replacement = false;
      }
      if (spec.precision_nested_replacement && spec.precision.has_value())
      {
        format_arg<OutputIt> prec_arg = args.get(spec.precision.value());
        size_t_extractor     ext;
        prec_arg.template visit<void>(ext);
        spec.precision                    = ext.value;
        spec.precision_nested_replacement = false;
      }
    }
 
    // Compute prefix/suffix padding sizes for alignment.
    // default_align_start: if true, NONE defaults to left-align (START); otherwise right-align (END).
    inline void compute_padding(size_t pad, spec_align_t align, bool default_align_start, size_t& prefix_size, size_t& suffix_size)
    {
      switch (align)
      {
        case spec_align_t::START:
          prefix_size = 0;
          suffix_size = pad;
          break;
        case spec_align_t::CENTER:
          prefix_size = pad / 2;
          suffix_size = pad - prefix_size;
          break;
        case spec_align_t::END:
          prefix_size = pad;
          suffix_size = 0;
          break;
        case spec_align_t::NONE:
        default:
          if (default_align_start)
          {
            prefix_size = 0;
            suffix_size = pad;
          }
          else
          {
            prefix_size = pad;
            suffix_size = 0;
          }
          break;
      }
    }
 
    template <typename OutputIt, typename Int>
    typename format_context<OutputIt>::iterator format_aligned_int(Int arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_num<private_format::counter_iterator, Int>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          compute_padding(pad, fmt_ctx.format_spec.align, false, prefix_size, suffix_size);
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
      private_format::fill<OutputIt>(it, prefix_size, fmt_ctx.format_spec.fill);
      private_format::format_num<OutputIt, Int>(it, arg, fmt_ctx.format_spec);
      private_format::fill<OutputIt>(it, suffix_size, fmt_ctx.format_spec.fill);
      return it;
    }
 
  #if ETL_USING_FORMAT_FLOATING_POINT
    template <typename OutputIt, typename Float>
    typename format_context<OutputIt>::iterator format_aligned_floating(Float arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      // For zero-padding ({:0Nf}), use '0' as fill and right-align (padding after sign)
      char_type fill_char = fmt_ctx.format_spec.fill;
      if (fmt_ctx.format_spec.zero && fmt_ctx.format_spec.align == spec_align_t::NONE)
      {
        fill_char = '0';
      }
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_floating<private_format::counter_iterator, Float>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          if (fmt_ctx.format_spec.zero && fmt_ctx.format_spec.align == spec_align_t::NONE)
          {
            // Zero-padding: all padding goes between sign and digits
            prefix_size = pad;
          }
          else
          {
            compute_padding(pad, fmt_ctx.format_spec.align, false, prefix_size, suffix_size);
          }
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
 
      if (fmt_ctx.format_spec.zero && fmt_ctx.format_spec.align == spec_align_t::NONE)
      {
        // Output sign first, then zero-fill, then the unsigned part
        bool sign = signbit(arg);
        if (sign || fmt_ctx.format_spec.sign != spec_sign_t::MINUS)
        {
          // Output the sign character
          char_type sc = '\0';
          if (sign)
          {
            sc = '-';
          }
          else
          {
            switch (fmt_ctx.format_spec.sign)
            {
              case spec_sign_t::PLUS: sc = '+'; break;
              case spec_sign_t::SPACE: sc = ' '; break;
              default: break;
            }
          }
          if (sc != '\0')
          {
            *it = sc;
            ++it;
          }
        }
        private_format::fill<OutputIt>(it, prefix_size, '0');
        // Format without sign (sign already emitted)
        format_spec_t no_sign_spec = fmt_ctx.format_spec;
        no_sign_spec.sign          = spec_sign_t::MINUS;
        Float abs_arg              = sign ? -arg : arg;
        private_format::format_floating<OutputIt, Float>(it, abs_arg, no_sign_spec);
      }
      else
      {
        private_format::fill<OutputIt>(it, prefix_size, fill_char);
        private_format::format_floating<OutputIt, Float>(it, arg, fmt_ctx.format_spec);
        private_format::fill<OutputIt>(it, suffix_size, fill_char);
      }
      return it;
    }
  #endif
 
    template <typename OutputIt>
    void format_string_view(OutputIt& it, etl::string_view arg, const format_spec_t& spec)
    {
      bool escaped = false;
      if (spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 's':
            // default output
            break;
          case '?':
            // escaped string
            escaped = true;
            break;
          default:
            // invalid type for string
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
      size_t limit = etl::numeric_limits<size_t>::max();
      if (spec.precision.has_value())
      {
        limit = spec.precision.value();
      }
 
      if (escaped)
      {
        format_plain_char(it, '"');
      }
      etl::string_view::const_iterator arg_it = arg.begin();
      while (arg_it != arg.cend() && limit > 0)
      {
        if (escaped)
        {
          format_escaped_char(it, *arg_it);
        }
        else
        {
          format_plain_char(it, *arg_it);
        }
        ++arg_it;
        --limit;
      }
      if (escaped)
      {
        format_plain_char(it, '"');
      }
    }
 
    template <typename OutputIt>
    typename format_context<OutputIt>::iterator format_aligned_string_view(etl::string_view arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_string_view<private_format::counter_iterator>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          compute_padding(pad, fmt_ctx.format_spec.align, true, prefix_size, suffix_size);
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
      private_format::fill<OutputIt>(it, prefix_size, fmt_ctx.format_spec.fill);
      private_format::format_string_view<OutputIt>(it, arg, fmt_ctx.format_spec);
      private_format::fill<OutputIt>(it, suffix_size, fmt_ctx.format_spec.fill);
      return it;
    }
 
    template <typename OutputIt>
    typename format_context<OutputIt>::iterator format_aligned_chars(const char* arg, format_context<OutputIt>& fmt_ctx)
    {
      return format_aligned_string_view<OutputIt>(etl::string_view(arg), fmt_ctx);
    }
 
    inline void check_char_spec(const format_spec_t& spec)
    {
      if ((!spec.type.has_value() || spec.type.value() == 'c' || spec.type.value() == '?')
          && (spec.sign != spec_sign_t::MINUS || spec.zero || spec.hash || spec.precision))
      {
        ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
      }
    }
 
    template <typename OutputIt>
    void format_char(OutputIt& it, char_type c, const format_spec_t& spec)
    {
      check_char_spec(spec);
      if (spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 'c':
            // default output
            format_plain_char(it, c);
            break;
          case '?':
            // escaped string
            format_plain_char(it, '\'');
            format_escaped_char(it, c);
            format_plain_char(it, '\'');
            break;
          case 'b':
          case 'B':
          case 'd':
          case 'o':
          case 'x':
          case 'X': private_format::format_num<OutputIt, unsigned int>(it, static_cast<unsigned int>(static_cast<unsigned char>(c)), spec); break;
          default:
            // invalid type for string
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
      else
      {
        format_plain_char(it, c);
      }
    }
 
    template <typename OutputIt>
    typename format_context<OutputIt>::iterator format_aligned_char(char_type arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_char<private_format::counter_iterator>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          // char type defaults to left-align, integer presentation defaults to right-align
          bool default_start =
            !fmt_ctx.format_spec.type.has_value() || fmt_ctx.format_spec.type.value() == 'c' || fmt_ctx.format_spec.type.value() == '?';
          compute_padding(pad, fmt_ctx.format_spec.align, default_start, prefix_size, suffix_size);
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
      private_format::fill<OutputIt>(it, prefix_size, fmt_ctx.format_spec.fill);
      private_format::format_char<OutputIt>(it, arg, fmt_ctx.format_spec);
      private_format::fill<OutputIt>(it, suffix_size, fmt_ctx.format_spec.fill);
      return it;
    }
 
    template <typename OutputIt>
    void format_bool(OutputIt& it, bool value, const format_spec_t& spec)
    {
      if (spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 's':
            // default output
            format_sequence(it, value ? "true" : "false");
            break;
          case 'b':
          case 'B':
          case 'd':
          case 'o':
          case 'x':
          case 'X': private_format::format_num<OutputIt, unsigned int>(it, static_cast<unsigned int>(static_cast<unsigned char>(value)), spec); break;
          default:
            // invalid type for string
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
      else
      {
        format_sequence(it, value ? "true" : "false");
      }
    }
 
    template <typename OutputIt>
    typename format_context<OutputIt>::iterator format_aligned_bool(bool arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_bool<private_format::counter_iterator>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          compute_padding(pad, fmt_ctx.format_spec.align, false, prefix_size, suffix_size);
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
      private_format::fill<OutputIt>(it, prefix_size, fmt_ctx.format_spec.fill);
      private_format::format_bool<OutputIt>(it, arg, fmt_ctx.format_spec);
      private_format::fill<OutputIt>(it, suffix_size, fmt_ctx.format_spec.fill);
      return it;
    }
 
    template <typename OutputIt>
    void format_pointer(OutputIt& it, const void* value, const format_spec_t& spec)
    {
      if (spec.type.has_value())
      {
        switch (spec.type.value())
        {
          case 'p':
          case 'P':
            format_sequence(it, spec.type.value() == 'p' ? "0x" : "0X");
            format_plain_num<OutputIt, uintptr_t>(it, reinterpret_cast<uintptr_t>(value), spec);
            break;
          default:
            // invalid type for string
            ETL_ASSERT_FAIL(ETL_ERROR(bad_format_string_exception));
        }
      }
      else
      {
        format_sequence(it, "0x");
        format_plain_num<OutputIt, uintptr_t>(it, reinterpret_cast<uintptr_t>(value), spec);
      }
    }
 
    template <typename OutputIt>
    typename format_context<OutputIt>::iterator format_aligned_pointer(const void* arg, format_context<OutputIt>& fmt_ctx)
    {
      size_t prefix_size = 0;
      size_t suffix_size = 0;
 
      if (fmt_ctx.format_spec.width)
      {
        private_format::counter_iterator counter;
        private_format::format_pointer<private_format::counter_iterator>(counter, arg, fmt_ctx.format_spec);
 
        if (counter.value() < fmt_ctx.format_spec.width.value())
        {
          size_t pad = fmt_ctx.format_spec.width.value() - counter.value();
          compute_padding(pad, fmt_ctx.format_spec.align, false, prefix_size, suffix_size);
        }
      }
 
      // actual output
      OutputIt it = fmt_ctx.out();
      private_format::fill<OutputIt>(it, prefix_size, fmt_ctx.format_spec.fill);
      private_format::format_pointer<OutputIt>(it, arg, fmt_ctx.format_spec);
      private_format::fill<OutputIt>(it, suffix_size, fmt_ctx.format_spec.fill);
      return it;
    }
  } // namespace private_format
 
  template <>
  struct formatter<int>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(int arg, format_context<OutputIt>& fmt_ctx)
    {
      if (fmt_ctx.format_spec.type.has_value() && fmt_ctx.format_spec.type.value() == 'c')
      {
        return private_format::format_aligned_char<OutputIt>(static_cast<private_format::char_type>(arg), fmt_ctx);
      }
      return private_format::format_aligned_int<OutputIt, int>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<unsigned int>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(unsigned int arg, format_context<OutputIt>& fmt_ctx)
    {
      if (fmt_ctx.format_spec.type.has_value() && fmt_ctx.format_spec.type.value() == 'c')
      {
        return private_format::format_aligned_char<OutputIt>(static_cast<private_format::char_type>(arg), fmt_ctx);
      }
      return private_format::format_aligned_int<OutputIt, unsigned int>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<long long int>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(long long int arg, format_context<OutputIt>& fmt_ctx)
    {
      if (fmt_ctx.format_spec.type.has_value() && fmt_ctx.format_spec.type.value() == 'c')
      {
        return private_format::format_aligned_char<OutputIt>(static_cast<private_format::char_type>(arg), fmt_ctx);
      }
      return private_format::format_aligned_int<OutputIt, long long int>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<unsigned long long int>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(unsigned long long int arg, format_context<OutputIt>& fmt_ctx)
    {
      if (fmt_ctx.format_spec.type.has_value() && fmt_ctx.format_spec.type.value() == 'c')
      {
        return private_format::format_aligned_char<OutputIt>(static_cast<private_format::char_type>(arg), fmt_ctx);
      }
      return private_format::format_aligned_int<OutputIt, unsigned long long int>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<char>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(private_format::char_type arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_char<OutputIt>(arg, fmt_ctx);
    }
  };
 
  #if ETL_USING_FORMAT_FLOATING_POINT
  template <>
  struct formatter<float>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(float arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_floating<OutputIt, float>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<double>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(double arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_floating<OutputIt, double>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<long double>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(long double arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_floating<OutputIt, long double>(arg, fmt_ctx);
    }
  };
  #endif
 
  template <>
  struct formatter<etl::string_view>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(etl::string_view arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_string_view<OutputIt>(arg, fmt_ctx);
    }
  };
 
  // string formatter
  template <>
  struct formatter<const char*>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(const char* arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_chars<OutputIt>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<bool>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(bool arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_bool<OutputIt>(arg, fmt_ctx);
    }
  };
 
  template <>
  struct formatter<const void*>
  {
    format_parse_context::iterator parse(format_parse_context& parse_ctx)
    {
      // unified parsing is done already in vformat_to()
      return parse_ctx.begin();
    }
 
    template <class OutputIt>
    typename format_context<OutputIt>::iterator format(const void* arg, format_context<OutputIt>& fmt_ctx)
    {
      return private_format::format_aligned_pointer<OutputIt>(arg, fmt_ctx);
    }
  };
 
  template <class OutputIt>
  OutputIt vformat_to(OutputIt out, etl::string_view fmt, format_args<OutputIt> args)
  {
    format_parse_context                     parse_context(fmt, args.size());
    format_context<OutputIt>                 fmt_context(out, args);
    private_format::format_visitor<OutputIt> v(parse_context, fmt_context);
 
    while (parse_context.begin() != parse_context.end())
    {
      const char c = *parse_context.begin();
      private_format::advance(parse_context);
      if (c == '{')
      {
        if (*parse_context.begin() == '{')
        {
          // escape sequence for literal '{'
          private_format::output<OutputIt>(fmt_context, c);
          private_format::advance(parse_context);
        }
        else
        {
          private_format::parse_format_spec<OutputIt>(parse_context, fmt_context);
 
          // Resolve nested replacement fields for width/precision
          private_format::resolve_nested_replacements<OutputIt>(fmt_context.format_spec, args);
 
          // Value index is always resolved in parse_format_spec
          size_t               index = fmt_context.format_spec.index.value();
          format_arg<OutputIt> arg   = args.get(index);
          arg.template visit<void>(v);
 
          ETL_ASSERT(*parse_context.begin() == '}', ETL_ERROR(bad_format_string_exception) /*"Closing brace missing"*/);
          if (parse_context.begin() != parse_context.end())
          {
            private_format::advance(parse_context);
          }
        }
      }
      else if (c == '}') // only matches here if } without { is found
      {
        ETL_ASSERT(*parse_context.begin() == '}', ETL_ERROR(bad_format_string_exception) /*"2nd closing brace missing on escaped closing brace"*/);
        // escape sequence for literal '}'
        private_format::output<OutputIt>(fmt_context, c);
        private_format::advance(parse_context);
      }
      else
      {
        private_format::output<OutputIt>(fmt_context, c);
      }
    }
 
    return fmt_context.out();
  }
 
  template <typename OutputIt, typename = etl::enable_if_t< !etl::is_base_of< etl::remove_reference<etl::istring>::type, OutputIt>::value>,
            class... Args>
  OutputIt format_to(OutputIt out, format_string<Args...> fmt, Args&&... args)
  {
    auto the_args{make_format_args<OutputIt>(args...)};
    return vformat_to(etl::move(out), fmt.get(), format_args<OutputIt>(the_args));
  }
 
  template <typename OutputIt, class WrapperIt = private_format::limit_iterator<OutputIt>, class... Args>
  OutputIt format_to_n(OutputIt out, size_t n, format_string<Args...> fmt, Args&&... args)
  {
    auto the_args{make_format_args<WrapperIt>(args...)};
    return vformat_to(WrapperIt(out, n), fmt.get(), format_args<WrapperIt>(the_args)).get();
  }
 
  // non std in the following, specific to etl
  template <class... Args>
  etl::istring::iterator format_to(etl::istring& out, format_string<Args...> fmt, Args&&... args)
  {
    etl::istring::iterator result = format_to_n(out.begin(), out.max_size(), fmt, etl::forward<Args>(args)...);
    out.uninitialized_resize(static_cast<size_t>(result - out.begin()));
    return result;
  }
 
  template <class... Args>
  size_t formatted_size(format_string<Args...> fmt, Args&&... args)
  {
    private_format::counter_iterator it;
    it = format_to(it, fmt, etl::forward<Args>(args)...);
    return it.value();
  }
} // namespace etl
 
#endif
 
#endif