Strings, Unicode, and standard containers

std::string owns contiguous bytes. string_view borrows bytes and carries a length; it need not end at a null terminator. A UTF-8 string's size counts code units, not Unicode code points or visible grapheme clusters. std::u32string stores code points but still does not make a combining sequence a single visible character.

std::string bytes="é";      // UTF-8 source execution encoding: usually two bytes
Text points=U"é";          // one char32_t code point
std::string embedded("a\0b",3); // size 3, despite the embedded NUL

Choose container semantics before choosing syntax. A vector owns contiguous elements and supports fast indexed access, but growth can invalidate pointers. A deque supports cheap growth at both ends, with different iterator invalidation rules. A list keeps node addresses stable across insertions but adds allocation and poor locality. An ordered map provides sorted keys and logarithmic lookup; an unordered map provides expected constant lookup without deterministic iteration order. operator[] on a map inserts a missing key, so it can mutate state during what looks like a read; use find/contains for read-only membership.

std::map<std::string,Int> counts;
auto found=counts.find("missing"); // no insertion
// counts["missing"] would create an entry containing zero.

UTF-8 decoding validates more than continuation-byte shape. Reject overlong encodings, surrogate code points, and values above U+10FFFF. A valid four-byte sequence is one output code point. An invalid byte sequence must not be silently accepted as text or passed to locale-dependent cctype functions. Always convert a byte to unsigned char before numeric bit inspection; signed char can otherwise sign-extend high bytes. A decoded string owns its code points and remains valid after the original bytes change. This exercise deliberately does not normalize, casefold, or combine graphemes.