Using a C++ user-defined literal to initialise an array

问题

I have a bunch of test vectors, presented in the form of hexadecimal strings:

MSG: 6BC1BEE22E409F96E93D7E117393172A
MAC: 070A16B46B4D4144F79BDD9DD04A287C
MSG: 6BC1BEE22E409F96E93D7E117393172AAE2D8A57
MAC: 7D85449EA6EA19C823A7BF78837DFADE

etc. I need to get these into a C++ program somehow, without too much editing required. There are various options:

Edit the test vectors by hand into the form 0x6B,0xC1,0xBE,...
Edit the test vectors by hand into the form "6BC1BEE22E409F96E93D7E117393172A" and write a function to convert that into a byte array at run time.
Write a program to parse the test vectors and output C++ code.

But the one I ended up using was:

User-defined literals,

because fun. I defined a helper class HexByteArray and a user-defined literal operator HexByteArray operator "" _$ (const char* s) that parses a string of the form "0xXX...XX", where XX...XX is an even number of hex digits. HexByteArray includes conversion operators to const uint8_t* and std::vector<uint8_t>. So now I can write e.g.

struct {
  std::vector<uint8_t> MSG ;
  uint8_t* MAC ;
  } Test1 = {
  0x6BC1BEE22E409F96E93D7E117393172A_$,
  0x070A16B46B4D4144F79BDD9DD04A287C_$
  } ;

Which works nicely. But now here is my question: Can I do this for arrays as well? For instance:

uint8_t MAC[16] = 0x070A16B46B4D4144F79BDD9DD04A287C_$ ;

or even

uint8_t MAC[] = 0x070A16B46B4D4144F79BDD9DD04A287C_$ ;

I can't see how to make this work. To initialise an array, I would seem to need an std::initializer_list. But as far as I can tell, only the compiler can instantiate such a thing. Any ideas?

Here is my code:

HexByteArray.h

#include <cstdint>
#include <vector>

class HexByteArray
  {
public:
  HexByteArray (const char* s) ;
  ~HexByteArray() { delete[] a ; }

  operator const uint8_t*() && { const uint8_t* t = a ; a = 0 ; return t ; }
  operator std::vector<uint8_t>() &&
    {
    std::vector<uint8_t> v ( a, a + len ) ;
    a = 0 ;
    return v ;
    }

  class ErrorInvalidPrefix { } ;
  class ErrorHexDigit { } ;
  class ErrorOddLength { } ;

private:
  const uint8_t* a = 0 ;
  size_t len ;
  } ;

inline HexByteArray operator "" _$ (const char* s)
  {
  return HexByteArray (s) ;
  }

HexByteArray.cpp

#include "HexByteArray.h"

#include <cctype>
#include <cstring>

HexByteArray::HexByteArray (const char* s)
  {
  if (s[0] != '0' || toupper (s[1]) != 'X') throw ErrorInvalidPrefix() ;
  s += 2 ;

  // Special case: 0x0_$ is an empty array (because 0x_$ is invalid C++ syntax)
  if (!strcmp (s, "0"))
    {
    a = nullptr ; len = 0 ;
    }
  else
    {
    for (len = 0 ; s[len] ; len++) if (!isxdigit (s[len])) throw ErrorHexDigit() ;
    if (len & 1) throw ErrorOddLength() ;
    len /= 2 ;
    uint8_t* t = new uint8_t[len] ;
    for (size_t i = 0 ; i < len ; i++, s += 2)
      sscanf (s, "%2hhx", &t[i]) ;
    a = t ;
    }
  }

回答1:

Use a numeric literal operator template, with the signature:

template <char...>
result_type operator "" _x();

Also, since the data is known at compile-time, we might as well make everything constexpr. Note that we use std::array instead of C-style arrays:

#include <cstdint>
#include <array>
#include <vector>

// Constexpr hex parsing algorithm follows:
struct InvalidHexDigit {};
struct InvalidPrefix {};
struct OddLength {};

constexpr std::uint8_t hex_value(char c)
{
    if ('0' <= c && c <= '9') return c - '0';
    // This assumes ASCII:
    if ('A' <= c && c <= 'F') return c - 'A' + 10;
    if ('a' <= c && c <= 'f') return c - 'a' + 10;
    // In constexpr-land, this is a compile-time error if execution reaches it:
    // The weird `if (c == c)` is to work around gcc 8.2 erroring out here even though
    // execution doesn't reach it.
    if (c == c) throw InvalidHexDigit{};
}

constexpr std::uint8_t parse_single(char a, char b)
{
    return (hex_value(a) << 4) | hex_value(b);
}

template <typename Iter, typename Out>
constexpr auto parse_hex(Iter begin, Iter end, Out out)
{
    if (end - begin <= 2) throw InvalidPrefix{};
    if (begin[0] != '0' || begin[1] != 'x') throw InvalidPrefix{};
    if ((end - begin) % 2 != 0) throw OddLength{};

    begin += 2;

    while (begin != end)
    {
        *out = parse_single(*begin, *(begin + 1));
        begin += 2;
        ++out;
    }

    return out;
}

// Make this a template to defer evaluation until later        
template <char... cs>
struct HexByteArray {
    static constexpr auto to_array()
    {
        constexpr std::array<char, sizeof...(cs)> data{cs...};

        std::array<std::uint8_t, (sizeof...(cs) / 2 - 1)> result{};

        parse_hex(data.begin(), data.end(), result.begin());

        return result;
    }

    constexpr operator std::array<std::uint8_t, (sizeof...(cs) / 2)>() const 
    {
        return to_array();
    }

    operator std::vector<std::uint8_t>() const
    {
        constexpr auto tmp = to_array();

        return std::vector<std::uint8_t>{tmp.begin(), tmp.end()};
    }
};

template <char... cs>
constexpr auto operator"" _$()
{
    static_assert(sizeof...(cs) % 2 == 0, "Must be an even number of chars");
    return HexByteArray<cs...>{};
}

Demo

Example usage:

auto data_array = 0x6BC1BEE22E409F96E93D7E117393172A_$ .to_array();
std::vector<std::uint8_t> data_vector = 0x6BC1BEE22E409F96E93D7E117393172A_$;

As a side note, $ in an identifier is actually a gcc extension, so it's non-standard C++. Consider using a UDL other than _$.

回答2:

This will make it

namespace detail{
template <std::size_t C> constexpr std::integral_constant<std::size_t, C> int_c{ };

template <char c>
class hex_decimal_t
{
    constexpr static std::uint8_t get_value() {
        constexpr std::uint8_t k = c - '0';
        if constexpr (k >= 0 && k <= 9) { return k; }
        else if constexpr (k >= 17 && k <= 22) { return k - 7;  }
        else if constexpr (k >= 49 && k <= 54) { return k - 39; }
        else { return std::uint8_t(-1); }
    }
public:
    static constexpr std::uint8_t value = get_value();
    constexpr operator auto() const{
        return value;
    }
};
template <char C> constexpr hex_decimal_t<C> hex_decimal{ };

template <bool B> using bool_type = std::integral_constant<bool, B>;

template <char... cs> struct is_valid_hex : std::false_type { };
template <char... cs> struct is_valid_hex<'0', 'x', cs...> : bool_type<((hex_decimal<cs> != std::uint8_t(-1)) && ...)>{};
template <char... cs> struct is_valid_hex<'0', 'X', cs...> : bool_type<((hex_decimal<cs> != std::uint8_t(-1)) && ...)>{};

template <std::size_t... Is>
constexpr auto expand_over(std::index_sequence<0, Is...>)
{
    return [](auto&& f) -> decltype(auto) {
        return decltype(f)(f)(int_c<Is>...);
    };
}

template <class T,class... F>
constexpr auto select(T, F&&... f) {
    return std::get<T{}>(std::forward_as_tuple(std::forward<F>(f)...));
}
}

template <char... ds>
constexpr auto operator "" _H()
{
    static_assert(detail::is_valid_hex<ds...>{} || sizeof...(ds) < 3, "Not a valid hex number");
    static_assert(!(sizeof...(ds) > 3 && sizeof...(ds) & 0x1), "Hex string must have even length");

    constexpr int Sz = sizeof...(ds);

    constexpr auto expand = detail::select(detail::int_c<(Sz > 3)>,
        [] { return detail::expand_over(std::make_index_sequence<2>{}); },
        [] { return detail::expand_over(std::make_index_sequence<Sz/2>{}); }
    )();

    if constexpr (Sz <= 3) {
        return expand([](auto... Is) {
            constexpr std::array digs{ds...};
            return std::array { (detail::hex_decimal<digs[2 * Is]>)... };
        });
    } else {
        return expand([](auto... Is) {
            constexpr std::array digs{ds...};
            return std::array { ((detail::hex_decimal<digs[2 * Is]> << 4) | detail::hex_decimal<digs[2 * Is + 1]>)... };
        });
    }
}

constexpr auto arr = 0x070A16B46B4D4144F79BDD9DD04A287C_H;
static_assert(arr.size() == 16);
static_assert(std::get<0>(arr) == 0x7);
static_assert(std::get<arr.size() - 1>(arr) == 0x7C);

Live demo

来源：https://stackoverflow.com/questions/53913800/using-a-c-user-defined-literal-to-initialise-an-array

标签

c++

initializer-list

user-defined-literals