2018年跨平台（Linux/Windows/macOS）使用C++处理Unicode的标准方案？

Great question—dealing with Unicode in C++ across Linux, Windows, and macOS can feel like navigating a mess, especially since the old Stack Overflow answers you found are now a decade out of date, and the tech landscape has shifted a lot.

Let’s break this down clearly, starting with the STL elephant in the room, then moving to practical solutions:

The STL’s Unicode Conundrum #

You’re right to be wary of wstring and codecvt_utf8. These tools have two big issues:

• Encoding inconsistency: On Windows, wchar_t is UTF-16, but on Linux/macOS it’s UTF-32—this alone creates cross-platform headaches.
• Deprecation and fragmentation: C17 deprecated parts of codecvt, and while C20 added std::u8string (a proper UTF-8 string type) and improved <regex>/<format> for Unicode, support across compilers is still spotty. Windows’ MSVC is playing catch-up here, so relying on modern STL features alone isn’t a safe cross-platform bet right now.

Practical Cross-Platform Solutions #

There’s no single "official standard" solution, but these are the most widely accepted options that cover all four of your requirements:

1. ICU (International Components for Unicode) #

This is the de facto industry standard for serious Unicode work across platforms. It checks every box for your needs:

• Reads/writes UTF-8 (and other encodings) to memory or disk seamlessly.
• Has a robust regex engine that fully supports Unicode characters, character codes, and lets you do replacements/formatting exactly as you need.
• Easily converts non-ASCII characters to ASCII+Unicode escape sequences (or any other format you need).
• Has pre-built packages for all three platforms:
  • Linux: Install via your package manager (e.g., libicu-dev on Debian/Ubuntu).
  • Windows: Use NuGet packages or pre-built binaries from the ICU project.
  • macOS: Install via Homebrew.

It’s a bit heavyweight, but if you need full Unicode support without reinventing the wheel, it’s the safest choice—used by major projects like Chrome, Firefox, and Qt.

2. Lightweight Library Combinations #

If ICU feels too big, you can pair a UTF-8 handling library with a Unicode-aware regex library:

• utfcpp: A tiny header-only library that handles UTF-8 validation, iteration, and basic conversions. It doesn’t do regex, though, so you’ll need to pair it with something like Boost.Regex (which has solid Unicode support) to cover your regex search/replace needs.
• unicorn-lib: Focuses more on Unicode character properties (like checking if a character is a letter, digit, etc.), so it’s less useful for your core IO/regex needs unless combined with other tools.

3. C++20+ STL (If You Can Target Modern Compilers) #

If your project is already using C++20 or later and you can assume your users have up-to-date compilers:

• std::u8string gives you a standard UTF-8 string type.
• The <format> library supports Unicode formatting, and <regex> has improved Unicode matching (via std::regex_constants::unicode).
• For IO, you can use std::fstream with UTF-8 strings (though Windows still has some quirks with console/file IO that may require extra work).

But again, this isn’t a universal solution yet—many projects still need to support older compilers, so proceed with caution.

Final Recommendation #

For a reliable, cross-platform solution that works today across all three platforms, ICU is your best bet. It’s well-maintained, fully covers all your requirements, and avoids the fragmentation of the STL’s Unicode tools.

内容的提问来源于stack exchange，提问作者Lfod