jes/src-c/core/Encoding.cc

#include "Encoding.h"

static inline bool
check_continuation_bytes(const uint8_t ** buffer,
        uint8_t continuation_bytes_to_check)
{
    for (uint8_t i = 0; i < continuation_bytes_to_check; i++)
    {
        if ((**buffer & 0xC0u) != 0x80u)
            return false;
        (*buffer)++;
    }
    return true;
}

Encoding::Type Encoding::detect_encoding(const uint8_t * buffer, size_t length)
{
    for (size_t i = 0u; i < length; i++)
    {
        const uint8_t c = *buffer;
        uint8_t continuation_bytes_to_check;
        if ((c & 0x80u) == 0u)
        {
            buffer++;
            continue;
        }
        else if ((c & 0xE0u) == 0xC0u)
        {
            continuation_bytes_to_check = 1u;
        }
        else if ((c & 0xF0u) == 0xE0u)
        {
            continuation_bytes_to_check = 2u;
        }
        else if ((c & 0xF8u) == 0xF0u)
        {
            continuation_bytes_to_check = 3u;
        }
        else if ((c & 0xFCu) == 0xF8u)
        {
            continuation_bytes_to_check = 4u;
        }
        else if ((c & 0xFEu) == 0xFCu)
        {
            continuation_bytes_to_check = 5u;
        }
        else
        {
            return CP_1252;
        }
        buffer++;
        if (((i + continuation_bytes_to_check) < length) &&
                check_continuation_bytes(&buffer, continuation_bytes_to_check))
        {
            i += continuation_bytes_to_check;
            continue;
        }
        return CP_1252;
    }

    return UTF_8;
}

uint8_t Encoding::num_bytes_in_code_point(Type type, const uint8_t * encoded)
{
    switch (type)
    {
    case UTF_8:
        {
            if ((*encoded & 0x80u) == 0u)
                return 1u;
            encoded++;
            uint8_t n = 1u;
            while ((*encoded++ & 0xC0u) == 0x80u)
                n++;
            return n;
        }
        break;
    case CP_1252:
        return 1u;
    }
    return 0u;
}

const uint8_t * Encoding::beginning_of_code_point(Type type, const uint8_t * encoded)
{
    switch (type)
    {
    case UTF_8:
        while ((*encoded & 0xC0u) == 0x80u)
            encoded--;
        return encoded;
    case CP_1252:
        return encoded;
    }

    return nullptr;
}

uint32_t Encoding::decode(Type type, const uint8_t * encoded, uint8_t * encoded_size)
{
    uint32_t result = 0u;
    uint8_t size = 0u;
    switch (type)
    {
    case UTF_8:
        {
            const uint_fast8_t c = *encoded;
            if ((c & 0x80u) == 0u)
            {
                result = c;
                size = 1u;
            }
            else
            {
                uint_fast8_t following_bytes = 0u;
                if ((c & 0xE0u) == 0xC0u)
                {
                    result = c & 0x1Fu;
                    following_bytes = 1u;
                }
                else if ((c & 0xF0u) == 0xE0u)
                {
                    result = c & 0x0Fu;
                    following_bytes = 2u;
                }
                else if ((c & 0xF8u) == 0xF0u)
                {
                    result = c & 0x07u;
                    following_bytes = 3u;
                }
                else if ((c & 0xFCu) == 0xF8u)
                {
                    result = c & 0x03u;
                    following_bytes = 4u;
                }
                else if ((c & 0xFEu) == 0xFCu)
                {
                    result = c & 0x01u;
                    following_bytes = 5u;
                }
                size = following_bytes + 1u;
                while (following_bytes-- > 0u)
                {
                    encoded++;
                    result <<= 6u;
                    result |= *encoded & 0x3Fu;
                }
            }
        }
        break;
    case CP_1252:
        /* TODO: map byte to code point */
        result = *encoded;
        size = 1u;
        break;
    }

    if (encoded_size != nullptr)
    {
        *encoded_size = size;
    }

    return result;
}

uint8_t Encoding::num_bytes_to_encode_code_point(uint32_t code_point, Type type)
{
    switch (type)
    {
    case UTF_8:
        if (code_point <= 0x7Fu)
        {
            return 1u;
        }
        else if (code_point <= 0x7FFu)
        {
            return 2u;
        }
        else if (code_point <= 0xFFFFu)
        {
            return 3u;
        }
        else if (code_point <= 0x1FFFFFu)
        {
            return 4u;
        }
        else if (code_point <= 0x3FFFFFFu)
        {
            return 5u;
        }
        else
        {
            return 6u;
        }
        break;
    case CP_1252:
        if (code_point <= 0xFFu)
        {
            return 1u;
        }
        else
        {
            return 0u;
        }
        break;
    }

    return 0u;
}

uint8_t Encoding::encode(uint32_t code_point, Type type, uint8_t * buffer)
{
    switch (type)
    {
    case UTF_8:
        {
            uint8_t size = num_bytes_to_encode_code_point(code_point, type);
            if (size == 1u)
            {
                *buffer = code_point;
            }
            else if (size > 1u)
            {
                *buffer = (0xFFu << (8u - size)) | (code_point >> (6u * (size - 1u)));
                for (uint8_t i = 0u, end = size - 1u; i < end; i++)
                {
                    buffer[end - i] = 0x80u | (code_point >> (6u * i) & 0x3Fu);
                }
            }
            return size;
        }
        break;
    case CP_1252:
        if (code_point <= 0xFFu)
        {
            *buffer = code_point;
            return 1u;
        }
        break;
    }

    return 0u;
}