'std::wstring_convert' to convert as much as possible (from a UTF8 file-read chunk)

问题

I am fetching text from a utf-8 text file, and doing it by chunks to increase performance.

std::ifstream.read(myChunkBuff_str, myChunkBuff_str.length())

Here is a more detailed example

I am getting around 16 thousand characters with each chunk. My next step is to convert this std::string into something that can allow me to work on these "complex characters" individually, thus converting that std::string into std::wstring.

I am using the following function for converting, taken from here:

#include <string>
#include <codecvt>
#include <locale>

std::string narrow (const std::wstring& wide_string)
{
    std::wstring_convert <std::codecvt_utf8 <wchar_t>, wchar_t> convert;
    return convert.to_bytes (wide_string);
}

std::wstring widen (const std::string& utf8_string)
{
    std::wstring_convert <std::codecvt_utf8 <wchar_t>, wchar_t> convert;
    return convert.from_bytes (utf8_string);
}

However, at its end of the chunk one of the Russian characters might be cut-off, and the conversion will fail, with an std::range_error exception.

For example, in UTF-8 "привет" takes 15 chars and "приве" takes 13 chars. So, if my chunk was hypothetically 14, the 'т' would be partially missing, and the conversion would throw exception.

Question:

How to detect these partially-loaded character? ('т' in this case) This would allow me to convert without it, and perhaps shift the next chunk a bit earlier than planned, to include this problematic 'т' next time?

I don't want to try or catch around these functions, as try/catch might slow me down the program. It also doesn't tell me "how much of character was missing for the conversion to actually succeed".

I know about wstring_convert::converted() but it's not really useful if my program crashes before I get to it

回答1:

You could do this using a couple of functions. UTF-8 has a way to detect the beginning of a multibyte character and (from the beginning) the size of the multibyte character.

So two functions:

// returns zero if this is the first byte of a UTF-8 char
// otherwise non-zero.
static unsigned is_continuation(char c)
{
    return (c & 0b10000000) && !(c & 0b01000000);
}

// if c is the *first* byte of a UTF-8 multibyte character, returns 
// the total number of bytes of the character.
static unsigned size(const unsigned char c)
{
    constexpr static const char u8char_size[] =
    {
          1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
        , 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
        , 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
        , 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
        , 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
        , 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2
        , 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2
        , 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3
        , 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 0, 0
    };
    return u8char_size[(unsigned char)c];
}

You could track back from the end of your buffer until is_continuation(c) is false. Then check if size(c) of the current UTF-8 char is longer than the end of the buffer.

Disclaimer - last time I looked these functions were working but have not used them in a while.

Edit: to add.

If you feel like doing th whole thing manually I may as well post the code to convert a UTF-8 multibyte character to a UTF-16 multibyte or a UTF-32 char.

UTF-32 Is easy:

// returns a UTF-32 char from a `UTF-8` multibyte
// character pointed to by cp
static char32_t char32(const char* cp)
{
    auto sz = size(*cp); // function above

    if(sz == 1)
        return *cp;

    char32_t c32 = (0b01111111 >> sz) & (*cp);

    for(unsigned i = 1; i < sz; ++i)
        c32 = (c32 << 6) | (cp[i] & 0b0011'1111);

    return c32;
}

UTF-16 Is a little more tricky:

// UTF-16 characters can be 1 or 2 characters wide...
using char16_pair = std::array<char16_t, 2>;

// outputs a UTF-16 char in cp16 from a `UTF-8` multibyte
// character pointed to by cp
//
// returns the number of characters in this `UTF-16` character
// (1 or 2).
static unsigned char16(const char* cp, char16_pair& cp16)
{
    char32_t c32 = char32(cp);

    if(c32 < 0xD800 || (c32 > 0xDFFF && c32 < 0x10000))
    {
        cp16[0] = char16_t(c32);
        cp16[1] = 0;
        return 1;
    }

    c32 -= 0x010000;

    cp16[0] = ((0b1111'1111'1100'0000'0000 & c32) >> 10) + 0xD800;
    cp16[1] = ((0b0000'0000'0011'1111'1111 & c32) >> 00) + 0xDC00;

    return 2;
}

来源：https://stackoverflow.com/questions/52338904/stdwstring-convert-to-convert-as-much-as-possible-from-a-utf8-file-read-chu

标签

c++

unicode

utf