Header string_width.hpp implements string width calculation as specified for std::format.

namespace string_width {

enum class surrogate_policy : bool {
    strict = false, lax = true
};

struct width_and_pos {
    int width;
    std::size_t pos;
};

template <typename CharT>
width_and_pos str_width_and_pos
    ( int width_limit
    , const CharT* str
    , std::size_t str_len
    , surrogate_policy surr_poli = surrogate_policy::strict ) noexcept

template <typename CharT>
int str_width
    ( int width_limit
    , const CharT* str
    , std::size_t str_len
    , surrogate_policy surr_poli = surrogate_policy::strict ) noexcept

} // namespace string_width

Function str_width_and_pos returns a value r such that r.pos is the size of the longest substring of str whose width is not greater than width_limit, and r.width is the width of such substring.

Function str_width returns only the width of such substring. However, it may be a little bit faster than str_width_and_pos in some situations.

str is expected to be encoded in UTF-8, UTF-16 or UTF-32 (depending on sizeof(CharT)).

The width of an invalid sequence is calculated as being the number of replacements characters that Unicode recommends to print when sanitizing such sequence. For example, in UTF-8, the sequences "\xE0\xA0" ( missing continuation byte ) and "\xE0\x9F\x80" ( overlong sequence ) are both invalid and are expected to be replaced by "\uFFFD" and "\uFFFD\uFFFD\uFFFD", respectively. Thus, theirs widths are calculated as 1 and 3, respectively.

surr_poli controls whether invalid surrogates are considered invalid. For example, in UTF-8, the sequence "\xED\xA0\x80" is invalid since it encodes U+D800, which is a surrogate, and it is expected to be replaced by "\uFFFD\uFFFD\uFFFD". So its width is 3 when surr_poli is surrogate_policy::strict. But if surr_poli is surrogate_policy::lax, then such sequence is considered valid and its width is 1.