Difference between revisions of "cpp/utility/compare/weak order"
m (link to ja) |
(describe as CPO, explain the weak order of floating-point types) |
||
Line 2: | Line 2: | ||
{{cpp/utility/navbar}} | {{cpp/utility/navbar}} | ||
{{dcl begin}} | {{dcl begin}} | ||
− | {{dcl header | compare}} | + | {{dcl header|compare}} |
− | {{dcl |since=c++20 | | + | {{dcl|since=c++20|1= |
− | template< class T > | + | inline namespace /* unspecified */ { |
− | constexpr std::weak_ordering weak_order( | + | inline constexpr /* unspecified */ weak_order = /* unspecified */; |
+ | } | ||
+ | }} | ||
+ | {{dcl h|Call signature}} | ||
+ | {{dcl|1= | ||
+ | template< class T, class U > | ||
+ | requires /* see below */ | ||
+ | constexpr std::weak_ordering weak_order(T&& t, U&& u) noexcept(/* see below */); | ||
}} | }} | ||
{{dcl end}} | {{dcl end}} | ||
Line 11: | Line 18: | ||
Compares two values using 3-way comparison and produces a result of type {{lc|std::weak_ordering}} | Compares two values using 3-way comparison and produces a result of type {{lc|std::weak_ordering}} | ||
− | + | Let {{tt|t}} and {{tt|u}} be expressions and {{tt|T}} and {{tt|U}} denote {{c|decltype((t))}} and {{c|decltype((u))}} respectively, {{c|std::weak_order(t, u)}} is expression-equivalent to: | |
− | + | * If {{c|1=std::is_same_v<std::decay_t<T>, std::decay_t<U>> == true}}: | |
− | + | ** {{c|std::weak_ordering(weak_order(t, u))}}, if it is a well-formed expression with overload resolution performed in a context that does not include a declaration of {{tt|std::weak_order}}, | |
− | * | + | ** otherwise, if {{tt|T}} is a floating-point type: |
− | + | *** if {{c|std::numeric_limits<T>::is_iec559}} is {{c|true}}, performs the weak ordering comparison of floating-point values (see below) and returns that result as a value of type {{lc|std::weak_ordering}}, | |
− | + | *** otherwise, yields a value of type {{lc|std::weak_ordering}} that is consistent with the ordering observed by {{tt|T}}'s comparison operators, | |
− | + | ** otherwise, {{c|1=std::weak_ordering(t <=> u)}} if it is well-formed, | |
− | + | ** other the expression is ill-formed, which can result in [[cpp/language/sfinae|substitution failure]] when it appears in the immediate context of a template instantiation. | |
+ | * Otherwise, {{c|std::weak_order(t, u)}} is ill-formed. | ||
− | + | {{cpp/expr-eq}} | |
− | {{ | + | {{cpp/ranges/cpo|std}} |
− | {{ | + | |
− | + | ||
− | + | ||
− | + | ||
− | + | ||
===Notes=== | ===Notes=== | ||
+ | |||
+ | ====Weak order of IEEE floating-point types==== | ||
+ | Let {{tt|x}} and {{tt|y}} be values of same IEEE floating-point type, and {{tt|weak_order_less(x, y)}} be the boolean result indicating if {{tt|x}} precedes {{tt|y}} in the strict weak order defined by the C++ standard. | ||
+ | |||
+ | * If neither {{tt|x}} nor {{tt|y}} is NaN, then {{c|1=weak_order_less(x, y) == true}} if and only if {{c|x < y}}, i.e. all representations of equal floating-point value are equivalent; | ||
+ | * If {{tt|x}} is negative NaN and {{tt|y}} is not negative NaN, then {{c|1=weak_order_less(x, y) == true}}; | ||
+ | * If {{tt|x}} is not positive NaN and {{tt|y}} is positive NaN, then {{c|1=weak_order_less(x, y) == true}}; | ||
+ | * If both {{tt|x}} and {{tt|y}} are NaNs with the same sign, then {{c|1=weak_order_less(x, y) {{!!}} weak_order_less(y, x) == false}}, i.e. all NaNs with the same sign are equivalent. | ||
===Example=== | ===Example=== |
Revision as of 00:19, 6 September 2019
Defined in header <compare>
|
||
inline namespace /* unspecified */ { inline constexpr /* unspecified */ weak_order = /* unspecified */; |
(since C++20) | |
Call signature |
||
template< class T, class U > requires /* see below */ |
||
Compares two values using 3-way comparison and produces a result of type std::weak_ordering
Let t
and u
be expressions and T
and U
denote decltype((t)) and decltype((u)) respectively, std::weak_order(t, u) is expression-equivalent to:
- If std::is_same_v<std::decay_t<T>, std::decay_t<U>> == true:
- std::weak_ordering(weak_order(t, u)), if it is a well-formed expression with overload resolution performed in a context that does not include a declaration of
std::weak_order
, - otherwise, if
T
is a floating-point type:- if std::numeric_limits<T>::is_iec559 is true, performs the weak ordering comparison of floating-point values (see below) and returns that result as a value of type std::weak_ordering,
- otherwise, yields a value of type std::weak_ordering that is consistent with the ordering observed by
T
's comparison operators,
- otherwise, std::weak_ordering(t <=> u) if it is well-formed,
- other the expression is ill-formed, which can result in substitution failure when it appears in the immediate context of a template instantiation.
- std::weak_ordering(weak_order(t, u)), if it is a well-formed expression with overload resolution performed in a context that does not include a declaration of
- Otherwise, std::weak_order(t, u) is ill-formed.
Contents |
Expression-equivalent
Expression e is expression-equivalent to expression f, if
- e and f have the same effects, and
- either both are constant subexpressions or else neither is a constant subexpression, and
- either both are potentially-throwing or else neither is potentially-throwing (i.e. noexcept(e) == noexcept(f)).
Customization point objects
The name std::weak_order
denotes a customization point object, which is a const function object of a literal semiregular
class type. For exposition purposes, the cv-unqualified version of its type is denoted as __weak_order_fn
.
All instances of __weak_order_fn
are equal. The effects of invoking different instances of type __weak_order_fn
on the same arguments are equivalent, regardless of whether the expression denoting the instance is an lvalue or rvalue, and is const-qualified or not (however, a volatile-qualified instance is not required to be invocable). Thus, std::weak_order
can be copied freely and its copies can be used interchangeably.
Given a set of types Args...
, if std::declval<Args>()... meet the requirements for arguments to std::weak_order
above, __weak_order_fn
models
- std::invocable<__weak_order_fn, Args...>,
- std::invocable<const __weak_order_fn, Args...>,
- std::invocable<__weak_order_fn&, Args...>, and
- std::invocable<const __weak_order_fn&, Args...>.
Otherwise, no function call operator of __weak_order_fn
participates in overload resolution.
Notes
Weak order of IEEE floating-point types
Let x
and y
be values of same IEEE floating-point type, and weak_order_less(x, y)
be the boolean result indicating if x
precedes y
in the strict weak order defined by the C++ standard.
- If neither
x
nory
is NaN, then weak_order_less(x, y) == true if and only if x < y, i.e. all representations of equal floating-point value are equivalent; - If
x
is negative NaN andy
is not negative NaN, then weak_order_less(x, y) == true; - If
x
is not positive NaN andy
is positive NaN, then weak_order_less(x, y) == true; - If both
x
andy
are NaNs with the same sign, then weak_order_less(x, y) || weak_order_less(y, x) == false, i.e. all NaNs with the same sign are equivalent.
Example
This section is incomplete Reason: no example |
See also
(C++20) |
the result type of 3-way comparison that supports all 6 operators and is not substitutable (class) |
(C++20) |
performs 3-way comparison and produces a result of type std::strong_ordering (customization point object) |
(C++20) |
performs 3-way comparison and produces a result of type std::partial_ordering (customization point object) |