parselink-old/include/parselink/msgpack/core/format.h

//-----------------------------------------------------------------------------
//     ___                      __ _       _
//    / _ \__ _ _ __ ___  ___  / /(_)_ __ | | __
//   / /_)/ _` | '__/ __|/ _ \/ / | | '_ \| |/ /
//  / ___/ (_| | |  \__ \  __/ /__| | | | |   <
//  \/    \__,_|_|  |___/\___\____/_|_| |_|_|\_\ .
//
//-----------------------------------------------------------------------------
// Author: Kurt Sassenrath
// Module: msgpack
//
// An implementation of the MessagePack binary format specification.
//
// Copyright (c) 2023 Kurt Sassenrath.
//
// License TBD.
//-----------------------------------------------------------------------------
#ifndef msgpack_format_849b5c5238d8212
#define msgpack_format_849b5c5238d8212

#include <bit>
#include <cstdint>
#include <span>
#include <string>
#include <string_view>
#include <type_traits>

namespace msgpack {

//-----------------------------------------------------------------------------
// Supporting types/tags for formats
//-----------------------------------------------------------------------------
struct nil {
    nil() = default;
    // This constructor is used by the reader implementation.
    nil(auto){};
};

struct invalid {};

using boolean = bool;

struct array_desc {
    constexpr array_desc() noexcept = default;
    constexpr array_desc(auto sz) noexcept
            : count(std::size_t(sz)){};
    // This constructor is implemented for use by reader
    constexpr array_desc(auto, auto sz) noexcept
            : count(std::size_t(sz)){};

    constexpr bool operator==(array_desc const& other) const noexcept {
        return count == other.count;
    }

    std::size_t count = 0;
};

struct map_desc {
    constexpr map_desc() noexcept = default;
    constexpr map_desc(auto sz) noexcept
            : count(std::size_t(sz)){};
    // This constructor is implemented for use by reader
    constexpr map_desc(auto, auto sz) noexcept
            : count(std::size_t(sz)){};
    std::size_t count = 0;
};

namespace format {

// Classification of the format type. Used by the token API to distinguish
// different formats.
enum class type : std::uint8_t {
    invalid,
    unsigned_int,
    signed_int,
    string,
    binary,
    nil,
    boolean,
    array,
    map,
    array_view,
    map_view,
};

// Flags that may control the behavior of readers/writers.
enum flag : std::uint8_t {
    apply_mask = 1,
    fixed_size = 2,
};

// MessagePack formats break down into one of the following schemes:
// Marker byte + Fixed-length value
// Marker byte + Fixed-length payload length + payload
//
// In addition, there are "fix" types that embed the literal value or the
// payload length within the marker byte.
enum payload : std::uint8_t { fixed, variable };

// Base structure for all MessagePack formats to inherit from.
struct base_format {
    constexpr static flag flags = {};
};

// Traits describing a particular format.
struct traits {
    std::uint8_t flags{};
    std::uint8_t size{};
    std::byte mask{};
    type token_type{};
    constexpr auto operator<=>(traits const&) const noexcept = default;
};

// "Sentinel" traits instance indicating no trait found
constexpr static traits no_traits{};

struct fixtype_format : base_format {};

template <class>
struct resolve_token_type {
    constexpr static type value = type::invalid;
};

template <std::integral T>
struct resolve_token_type<T> {
    constexpr static type value =
            std::is_signed_v<T> ? type::signed_int : type::unsigned_int;
};

template <>
struct resolve_token_type<std::string_view> {
    constexpr static type value = type::string;
};

template <>
struct resolve_token_type<std::span<std::byte const>> {
    constexpr static type value = type::binary;
};

template <>
struct resolve_token_type<bool> {
    constexpr static type value = type::boolean;
};

template <>
struct resolve_token_type<nil> {
    constexpr static type value = type::nil;
};

template <>
struct resolve_token_type<invalid> {
    constexpr static type value = type::invalid;
};

template <>
struct resolve_token_type<array_desc> {
    constexpr static type value = type::array;
};

template <>
struct resolve_token_type<map_desc> {
    constexpr static type value = type::map;
};

template <typename T>
constexpr static auto resolve_type_v = resolve_token_type<T>::value;

// The library's representation of certain data types does not always
// match the underlying data serialized in the message. For example,
// arrays and maps are encoded as a marker byte + fixed length value, but
// msgpack will present them wrapped in a structure for stronger type
// semantics.

template <std::uint8_t Marker, typename First, typename Value = First>
struct format : base_format {
    using first_type = First;
    using value_type = Value; // Can be overridden
    constexpr static std::byte marker{Marker};
    constexpr static payload payload_type{payload::fixed};
    constexpr static std::uint8_t flags{};
};

template <std::uint8_t Marker, std::uint8_t Mask, typename First = std::uint8_t>
struct fixtype : fixtype_format {
    using first_type = First;
    using value_type = first_type; // Can be overridden
    constexpr static std::byte marker{Marker};
    constexpr static std::byte mask{Mask};
    constexpr static payload payload_type{payload::fixed};
    constexpr static auto flags{flag::apply_mask};
};

/*
 * Integral types
 */

// Positive/negative fixint represent the literal value specified, do not
// need to mask it off.
struct positive_fixint : fixtype<0x00, 0x7f> {
    constexpr static flag flags{};
};

struct negative_fixint : fixtype<0xe0, 0x1f, std::int8_t> {
    constexpr static flag flags{};
};

struct uint8 : format<0xcc, std::uint8_t> {};

struct uint16 : format<0xcd, std::uint16_t> {};

struct uint32 : format<0xce, std::uint32_t> {};

struct uint64 : format<0xcf, std::uint64_t> {};

struct int8 : format<0xd0, std::int8_t> {};

struct int16 : format<0xd1, std::int16_t> {};

struct int32 : format<0xd2, std::int32_t> {};

struct int64 : format<0xd3, std::int64_t> {};

/*
 * Other primitive types
 */
struct nil : fixtype<0xc0, 0x00> {
    using value_type = msgpack::nil;
    constexpr static flag flags{flag::fixed_size | flag::apply_mask};
};

struct invalid : fixtype<0xc1, 0x00> {
    using value_type = msgpack::invalid;
    constexpr static flag flags{flag::fixed_size | flag::apply_mask};
};

struct boolean : fixtype<0xc2, 0x01> {
    using value_type = bool;
    constexpr static flag flags{flag::fixed_size | flag::apply_mask};
};

/*
 * Maps
 */

template <typename Fmt>
struct map_format : Fmt {
    using value_type = map_desc;
};

struct fixmap : map_format<fixtype<0x80, 0x0f>> {};

struct map16 : map_format<format<0xde, std::uint16_t>> {};

struct map32 : map_format<format<0xdf, std::uint32_t>> {};

/*
 * Arrays
 */

template <typename Fmt>
struct array_format : Fmt {
    using value_type = array_desc;
};

struct fixarray : array_format<fixtype<0x90, 0x0f>> {};

struct array16 : array_format<format<0xdc, std::uint16_t>> {};

struct array32 : array_format<format<0xdd, std::uint32_t>> {};

/*
 * Strings
 */

template <typename Fmt>
struct string_format : Fmt {
    using value_type = std::string_view;
    constexpr static auto payload_type = payload::variable;
};

struct fixstr : string_format<fixtype<0xa0, 0x1f>> {};

struct str8 : string_format<format<0xd9, std::uint8_t>> {};

struct str16 : string_format<format<0xda, std::uint16_t>> {};

struct str32 : string_format<format<0xdb, std::uint32_t>> {};

/*
 * Binary arrays
 */

template <typename Fmt>
struct bin_format : Fmt {
    using value_type = std::span<std::byte const>;
    constexpr static auto payload_type = payload::variable;
};

struct bin8 : bin_format<format<0xc4, std::uint8_t>> {};

struct bin16 : bin_format<format<0xc5, std::uint16_t>> {};

struct bin32 : bin_format<format<0xc6, std::uint32_t>> {};

/*
 * Extension types, not yet supported.
 */

template <typename Fmt>
struct unimplemented_format : Fmt {};

struct fixext1 : unimplemented_format<fixtype<0xd4, 0x00>> {};

struct fixext2 : unimplemented_format<fixtype<0xd5, 0x00>> {};

struct fixext4 : unimplemented_format<fixtype<0xd6, 0x00>> {};

struct fixext8 : unimplemented_format<fixtype<0xd7, 0x00>> {};

struct fixext16 : unimplemented_format<fixtype<0xd8, 0x00>> {};

struct ext8 : unimplemented_format<format<0xc7, std::uint8_t>> {};

struct ext16 : unimplemented_format<format<0xc8, std::uint16_t>> {};

struct ext32 : unimplemented_format<format<0xc9, std::uint32_t>> {};

template <typename T>
struct unimplemented : std::false_type {};

template <typename Fmt>
struct unimplemented<unimplemented_format<Fmt>> : std::true_type {};

namespace detail {
// Simple typelist for looking up compatible types.
// TODO: Use traits to generate the compatibility list automatically?
template <std::size_t I, typename T>
struct type_list_entry {
    using type = T;
};

template <typename...>
struct type_list_impl;

template <std::size_t... Is, typename... Ts>
struct type_list_impl<std::index_sequence<Is...>, Ts...>
        : type_list_entry<Is, Ts>... {
    static constexpr auto size = sizeof...(Ts);
};

template <typename T>
struct typelist_lookup_tag {
    using type = T;
};

template <std::size_t I, typename T>
typelist_lookup_tag<T> typelist_lookup(type_list_entry<I, T> const&);

template <std::size_t I, typename List>
using type_list_at =
        typename decltype(typelist_lookup<I>(std::declval<List>()))::type;

template <typename... Ts>
using type_list =
        detail::type_list_impl<std::make_index_sequence<sizeof...(Ts)>, Ts...>;

template <typename, typename, template <typename> typename>
struct filter_type_list;

template <typename... Ts, template <typename> typename Predicate>
struct filter_type_list<type_list<>, type_list<Ts...>, Predicate> {
    using type = type_list<Ts...>;
};

template <typename Head, typename... Tail, typename... Ts,
        template <typename> typename Predicate>
struct filter_type_list<type_list<Head, Tail...>, type_list<Ts...>, Predicate> {
    using type = typename std::conditional_t<Predicate<Head>::value,
            filter_type_list<type_list<Tail...>, type_list<Ts..., Head>,
                    Predicate>,
            filter_type_list<type_list<Tail...>, type_list<Ts...>,
                    Predicate>>::type;
};

template <typename List, template <typename> typename Predicate>
using filter = filter_type_list<List, type_list<>, Predicate>;

template <typename List, template <typename> typename Predicate>
using filter_t = typename filter<List, Predicate>::type;
} // namespace detail
} // namespace format

template <typename T>
concept format_type = std::is_base_of_v<format::base_format, T>;

template <typename T>
concept is_fixtype = std::is_base_of_v<format::fixtype_format, T>;

namespace format {
// This template instantiates a format::traits object for a given format.
// This should only occur once per format type, and should happen at
// compile time.
//
// This isn't the prettiest way to go about it.
template <format_type Format>
inline traits const& get_traits() noexcept {
    constexpr static auto inst = [] {
        // Fixtypes define the size within the identifier byte, so the
        // trait size must be zero.
        auto size =
                is_fixtype<Format> ? 0 : sizeof(typename Format::first_type);

        traits inst{.flags = Format::flags,
                .size = std::uint8_t(size),
                .token_type = resolve_type_v<typename Format::value_type>};
        // Add in the fixed_size flag for integral types.
        if constexpr (std::is_integral_v<typename Format::value_type>) {
            inst.flags |= flag::fixed_size;
        }
        if constexpr (is_fixtype<Format>) {
            inst.mask = Format::mask;
        }
        return inst;
    }();
    return inst;
};
} // namespace format

template <format_type... Formats>
using format_list = format::detail::type_list<Formats...>;

template <std::size_t I, typename List>
using format_list_at = format::detail::type_list_at<I, List>;

} // namespace msgpack

#endif // oh_msgpack_format_849b5c5238d8212