3.0

This is a major release with some breaking changes

Changed

• C++20 or higher is now required for compilation. In particular, the following C++20 features must be available:
  • Ranges support in standard library (__cpp_lib_ranges >= 201911)
  • Three-way comparison (spaceship operator)
  • char8_t type
  • std::endian support in standard library (__cpp_lib_endian >= 201907)
  • Minimal compilers known to work include: GCC 12, Clang 16, Apple Clang 15.4 and MSVC 17.6.
• The library has been range-ified.
  • All methods that used to accept iterator pairs now take iterator/sentinel pairs.
  • All these methods now also have overloads that accept ranges
  • Existing informal ranges (sys_string::char_access, sys_string::utf_view, etc.) are now
    formal ranges or views.
  • As part of the above sys_string::utfX_view classes has been renamed to sys_string::utfX_access (because they are
    not formally views as defined by standard library). The old names have been retained for compatibility but annotated
    as deprecated. Note that sys_string_builder::utf_view remains under the same name since it is a view.
  • Breaking change: as part of the above change the sys_string::utf_access and sys_string_builder::utf_view now
    return distinct iterators and sentinels (that is they no longer satisfy std::ranges::common_range concept).
    You will need to use ranges algorithms with their iterators (e.g. std::ranges::for_each rather than std::for_each).
  • The global utf_view template has been split into two: utf_ref_view that takes underlying range by reference (similar
    to std::ref_view) and utf_owning_view that owns a movable underlying range (similar to std::owning_view). These
    are automatically produced by as_utf range adapter closures (see below in Added section)
  • Breaking change: the non-standard Cursor classes has been removed.
• The library has been concept-ified.
  • Most templated library calls now have concepts checks that validate their argument types.
  • Primitive std::enable_if used before have been subsumed by these and removed.
• Unicode data used for case folding and whitespace detection has been updated to version 16.0.0

Added

• sys_string_t can now be +-ed with any forward range of any type of character (including C strings and std::string).
  This results in a the same optimized addition as when adding sys_string_t objects.
• sys_string_t objects can now be formatted via std::format (if available in your library). On platforms
  where wchar_t is UTF-16 or UTF-32 you can also use wide character formatting.
• sys_string_t::std_format method. This formats a new sys_string_t (similar to the existing sys_string_t::format)
  but uses std::format machinery and formatting string syntax.
• Range adapter closures: as_utf8, as_utf18, as_utf32 and generic as_utf<encoding> .
  • These can be used to create utf_ref_view/utf_owning_view from any range/view. For example as_utf16(std::string("abc"))
  • If you library supports custom adapter closures (usually __cpp_lib_ranges >= 202202L) they can be used in
    view pipelines like std::string("abc") | as_utf16 | std::views::take(2) etc.

Fixed

• Printing sys_string_t objects into std::ostream (and std::wostream if available) now functions correctly in presence
  of stream formatting flags. Flags are currently ignored. This might change in a future version.
• Printing/formatting sys_string_t objects that use char storage type now does not perform sanitizing transcoding. The content
  of the string is printed as-is. This allows faithful round-tripping and support for invalid Unicode for those scenarios. Similar
  behavior applies to wchar_t on platform where it is UTF-16 or UTF-32.
• operator<< no longer pollutes global namespace