Here is what I have done to try to capture the changes to C++17 from Jacksonville.
The vast majority of the library proposals include a proposed name for the feature-test macro. (Yay!) They are all included in the table. But none of the language proposals did, and a few of the library proposals didn't say anything.
I have added some proposals of my own, marked with a question mark. Note that for range-based "for", constexpr lambdas, and fold expressions, the proposal is to bump an existing macro. (The target of the link from the new value is in SD-6; it dangles in this document.)
For the enum class value change, the "design mistake", and non-const data for string, the proposal is that no macro is needed; rationale is included.
The fundamentals TS (and to a lesser degree the parallelism TS) make it not obvious what to do about keeping the table sorted by primary section number.
Most of the document links are broken now, but should start working when the post-meeting mailing is posted.
Doc. No. | Title | Primary Section | Macro Name | Value | Header |
---|---|---|---|---|---|
P0245R1 | Hexadecimal floating literals for C++ | 2.13 | __cpp_hex_float |
201603 | predefined |
P0018R3 | Lambda Capture of *this by Value as [=,*this] | 5.1 | __cpp_capture_this ?__cpp_capture_star_this ? |
201603 | predefined |
P0170R1 | Wording for Constexpr Lambda | 5.1 | __cpp_constexpr |
201603 | predefined |
P0184R0 | Generalizing the Range-Based For Loop | 6.5 | __cpp_range_based_for |
201603 | predefined |
P0188R1 | Wording for [[fallthrough]] attribute | 7.6 | __has_cpp_attribute(fallthrough) |
201603 | predefined |
P0189R1 | Wording for [[nodiscard]] attribute | 7.6 | __has_cpp_attribute(nodiscard) |
201603 | predefined |
P0212R1 | Wording for [[maybe_unused]] attribute | 7.6 | __has_cpp_attribute(maybe_unused) |
201603 | predefined |
P0017R1 | Extension to aggregate initialization | 8.5 | __cpp_aggregate_bases |
201603 | predefined |
P0138R2 | Construction Rules for enum class Values | 8.5 | none ? | ||
P0036R0 | Unary Folds and Empty Parameter Packs | 14.5 | __cpp_fold_expressions |
201603 | predefined |
P0154R1 | constexpr std::hardware_{constructive,destructive}_interference_size | 18.6 | __cpp_lib_hardware_interference_size |
201603 | <new> |
LWG2296 | std::addressof should be constexpr | 20.7 | __cpp_lib_addressof_constexpr ? |
201603 | <memory> |
P0033R1 | Re-enabling shared_from_this | 20.8 | __cpp_lib_enable_shared_from_this
|
201603 | <memory> |
P0005R4 | Adopt not_fn from Library Fundamentals 2 for C++17 | 20.9 | __cpp_lib_not_fn
|
201603 | <function> |
P0253R1 | Fixing a design mistake in the searchers interface in Library Fundamentals | 20.9 | none | ||
P0185R1 | Adding [nothrow-]swappable traits | 20.10 | __cpp_lib_is_swappable
|
201603 | <type_traits> |
P0077R2 | is_callable, the missing INVOKE related trait | 20.10 | __cpp_lib_is_callable
|
201603 | <type_traits> |
P0220R1 | Adopt Library Fundamentals V1 TS Components for C++17 | 20.4 | __cpp_lib_apply |
201603 | <tuple> |
20.8 | __cpp_lib_shared_ptr_arrays |
201603 | <memory> |
||
20.9 | __cpp_lib_boyer_moore_searching |
201603 | <functional> |
||
20 |
__has_include(<optional>) __has_include(<any>) __has_include(<string_view>) __has_include(<memory_resource>)
|
1 | predefined | ||
P0272R1 | Give 'std::string' a non-const '.data()' member function | 21.4 | none ? | ||
P0031R0 | A Proposal to Add Constexpr Modifiers to reverse_iterator, move_iterator, array and Range Access | 24.3 | __cpp_lib_array_constexpr
|
201603 | <iterator> |
P0025R0 | An algorithm to "clamp" a value between a pair of boundary values | 25.4 | __cpp_lib_clamp
|
201603 | <algorithm> |
P0024R2 | The Parallelism TS Should be Standardized | 25, 26 | __cpp_lib_parallel_algorithm
|
201603 | <algorithm> <numeric> |
18 | __has_include(<exception_list>) |
1 | predefined | ||
20 | __has_include(<execution_policy>) |
1 | predefined | ||
P0030R1 | Proposal to Introduce a 3-Argument Overload to std::hypot | 26.8 | __cpp_lib_hypot
|
201603 | <cmath> |
P0226R1 | Mathematical Special Functions for C++17 | 26.8 | __cpp_lib_math_special_functions
|
201603 | <cmath> |
P0218R1 | Adopt the File System TS for C++17 | 27.10 | __has_include(<filesystem>) |
1 | predefined |
__cpp_lib_filesystem
|
201603 | <filesystem> |
|||
P0152R1 | constexpr atomic<T>::is_always_lock_free | 29.5 | __cpp_lib_atomic_is_always_lock_free |
201603 | <atomic> |
This doesn't provide any new functionality; it just makes some kinds of code simpler to write. Code that needs to be portable to implementations lacking this feature would be still more complicated if it tried to use the new feature when available. So a macro for this feature would not seem to be justified.
Many thanks to Daniel for sending me this example – unsolicited!
Example:
template<class T> void my_swap(T& x, T& y) #if __cpp_lib_is_swappable noexcept(std::is_swappable_v<T>); // Covers all Lvalue Swappable cases #else noexcept( std::is_nothrow_move_constructible_v<T> && std::is_nothrow_move_assignable_v<T> ); #endif
The fundamentals TS proposed a macro for each added component declared in a new header, but the parallelism TS added new headers, with only a single over-arching macro. Requiring a version-test macro for every brand-new header seems to be a bit excessive.
This probably doesn't belong here long-term; it's just for short-term confirmation. Omitted:
This is a fix that is applied to the searchers interface at the same time that it is incorporated into the standard. The unfixed state exists only in the TS, which has its own macro name and value. The new macro for the searchers interface in the standard, with its new value, will be enough to indicate that this fix is applied.
This doesn't provide any new functionality; it just makes some kinds of code simpler to write. Code that needs to be portable to implementations lacking this feature would be still more complicated if it tried to use the new feature when available. So a macro for this feature would not seem to be justified.
Example:
struct optional { bool b; T t; constexpr T *get() { return b ? #if __cpp_lib_addressof_constexpr std::addressof(t) #else &t #endif : nullptr; } };
It is more important for get
to be
constexpr
than for it to work
if there is an overloaded operator&
,
so addressof
is used
only if it would actually work in a constant expression.