The Guidelines Support Library (GSL) interface is very lightweight and exposed via a header-only library. This document attempts to document all of the headers and their exposed classes and functions.
Types and functions are exported in the namespace gsl
.
See GSL: Guidelines support library
This header contains some common algorithms that have been wrapped in GSL safety features.
template <class SrcElementType, std::size_t SrcExtent, class DestElementType,
std::size_t DestExtent>
void copy(span<SrcElementType, SrcExtent> src, span<DestElementType, DestExtent> dest);
This function copies the content from the src
span
to the dest
span
. It Expects
that the destination span
is at least as large as the source span
.
This header contains some macros used for contract checking and suppressing code analysis warnings.
This macro can be used to suppress a code analysis warning.
The core guidelines request tools that check for the rules to respect suppressing a rule by writing
[[gsl::suppress(tag)]]
or [[gsl::suppress(tag, justification: "message")]]
.
Clang does not use exactly that syntax, but requires tag
to be put in double quotes [[gsl::suppress("tag")]]
.
For portable code you can use GSL_SUPPRESS(tag)
.
This macro can be used for expressing a precondition. If the precondition is not held, then std::terminate
will be called.
See I.6: Prefer Expects()
for expressing preconditions
This macro can be used for expressing a postcondition. If the postcondition is not held, then std::terminate
will be called.
See I.8: Prefer Ensures()
for expressing postconditions
This header contains the definition of a byte type, implementing std::byte
before it was standardized into C++17.
If GSL_USE_STD_BYTE
is defined to be 1
, then gsl::byte
will be an alias to std::byte
.
If GSL_USE_STD_BYTE
is defined to be 0
, then gsl::byte
will be a distinct type that implements the concept of byte.
If GSL_USE_STD_BYTE
is not defined, then the header file will check if std::byte
is available (C++17 or higher). If yes,
gsl::byte
will be an alias to std::byte
, otherwise gsl::byte
will be a distinct type that implements the concept of byte.
⚠ Take care when linking projects that were compiled with different language standards (before C++17 and C++17 or higher).
If you do so, you might want to #define GSL_USE_STD_BYTE 0
to a fixed value to be sure that both projects use exactly
the same type. Otherwise you might get linker errors.
See SL.str.5: Use std::byte
to refer to byte values that do not necessarily represent characters
template <class IntegerType, class = std::enable_if_t<std::is_integral<IntegerType>::value>>
constexpr byte& operator<<=(byte& b, IntegerType shift) noexcept;
template <class IntegerType, class = std::enable_if_t<std::is_integral<IntegerType>::value>>
constexpr byte operator<<(byte b, IntegerType shift) noexcept;
template <class IntegerType, class = std::enable_if_t<std::is_integral<IntegerType>::value>>
constexpr byte& operator>>=(byte& b, IntegerType shift) noexcept;
template <class IntegerType, class = std::enable_if_t<std::is_integral<IntegerType>::value>>
constexpr byte operator>>(byte b, IntegerType shift) noexcept;
Left or right shift a byte
by a given number of bits.
constexpr byte& operator|=(byte& l, byte r) noexcept;
constexpr byte operator|(byte l, byte r) noexcept;
Bitwise "or" of two byte
s.
constexpr byte& operator&=(byte& l, byte r) noexcept;
constexpr byte operator&(byte l, byte r) noexcept;
Bitwise "and" of two byte
s.
constexpr byte& operator^=(byte& l, byte r) noexcept;
constexpr byte operator^(byte l, byte r) noexcept;
Bitwise xor of two byte
s.
constexpr byte operator~(byte b) noexcept;
Bitwise negation of a byte
. Flips all bits. Zeroes become ones, ones become zeroes.
template <class IntegerType, class = std::enable_if_t<std::is_integral<IntegerType>::value>>
constexpr IntegerType to_integer(byte b) noexcept;
Convert the given byte
value to an integral type.
template <typename T>
constexpr byte to_byte(T t) noexcept;
Convert the given value to a byte
. The template requires T
to be an unsigned char
so that no data loss can occur.
If you want to convert an integer constant to a byte
you probably want to call to_byte<integer constant>()
.
template <int I>
constexpr byte to_byte() noexcept;
Convert the given value I
to a byte
. The template requires I
to be in the valid range 0..255 for a gsl::byte
.
This header is a convenience header that includes all other GSL headers.
Since <narrow>
requires exceptions, it will only be included if exceptions are enabled.
This header contains utility functions and classes, for narrowing casts, which require exceptions. The narrowing-related utilities that don't require exceptions are found inside util.
gsl::narrowing_error
is the exception thrown by gsl::narrow
when a narrowing conversion fails. It is derived from std::exception
.
gsl::narrow<T>(x)
is a named cast that does a static_cast<T>(x)
for narrowing conversions with no signedness promotions.
If the argument x
cannot be represented in the target type T
, then the function throws a gsl::narrowing_error
(e.g., narrow<unsigned>(-42)
and narrow<char>(300)
throw).
Note: compare gsl::narrow_cast
in header util.
See ES.46: Avoid lossy (narrowing, truncating) arithmetic conversions and ES.49: If you must use a cast, use a named cast
This header contains some pointer types.
See GSL.view
gsl::unique_ptr
is an alias to std::unique_ptr
.
See GSL.owner: Ownership pointers
gsl::shared_ptr
is an alias to std::shared_ptr
.
See GSL.owner: Ownership pointers
gsl::owner<T>
is designed as a safety mechanism for code that must deal directly with raw pointers that own memory. Ideally such code should be restricted to the implementation of low-level abstractions. gsl::owner
can also be used as a stepping point in converting legacy code to use more modern RAII constructs such as smart pointers.
T
must be a pointer type (std::is_pointer<T>
).
A gsl::owner<T>
is a typedef to T
. It adds no runtime overhead whatsoever, as it is purely syntactic and does not add any runtime checks. Instead, it serves as an annotation for static analysis tools which check for memory safety, and as a code comprehension guide for human readers.
See Enforcement section of C.31: All resources acquired by a class must be released by the class’s destructor.
gsl::not_null<T>
restricts a pointer or smart pointer to only hold non-null values. It has no size overhead over T
.
The checks for ensuring that the pointer is not null are done in the constructor. There is no overhead when retrieving or dereferencing the checked pointer.
When a nullptr check fails, std::terminate
is called.
See F.23: Use a not_null<T>
to indicate that “null” is not a valid value
template <typename U, typename = std::enable_if_t<std::is_convertible<U, T>::value>>
constexpr not_null(U&& u);
template <typename = std::enable_if_t<!std::is_same<std::nullptr_t, T>::value>>
constexpr not_null(T u);
Constructs a gsl_owner<T>
from a pointer that is convertible to T
or that is a T
. It Expects
that the provided pointer is not == nullptr
.
template <typename U, typename = std::enable_if_t<std::is_convertible<U, T>::value>>
constexpr not_null(const not_null<U>& other);
Constructs a gsl_owner<T>
from another gsl_owner
where the other pointer is convertible to T
. It Expects
that the provided pointer is not == nullptr
.
not_null(const not_null& other) = default;
not_null& operator=(const not_null& other) = default;
Copy construction and assignment.
not_null(std::nullptr_t) = delete;
not_null& operator=(std::nullptr_t) = delete;
Construction from std::nullptr_t
and assignment of std::nullptr_t
are explicitly deleted.
not_null& operator++() = delete;
not_null& operator--() = delete;
not_null operator++(int) = delete;
not_null operator--(int) = delete;
not_null& operator+=(std::ptrdiff_t) = delete;
not_null& operator-=(std::ptrdiff_t) = delete;
Explicitly deleted operators. Pointers point to single objects (I.13: Do not pass an array as a single pointer), so don't allow these operators.
constexpr details::value_or_reference_return_t<T> get() const;
constexpr operator T() const { return get(); }
Get the underlying pointer.
constexpr decltype(auto) operator->() const { return get(); }
constexpr decltype(auto) operator*() const { return *get(); }
Dereference the underlying pointer.
void operator[](std::ptrdiff_t) const = delete;
Array index operator is explicitly deleted. Pointers point to single objects (I.13: Do not pass an array as a single pointer), so don't allow treating them as an array.
template <class T>
auto make_not_null(T&& t) noexcept;
Creates a gsl::not_null
object, deducing the target type from the type of the argument.
template <class T, class U>
auto operator==(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() == rhs.get()))
-> decltype(lhs.get() == rhs.get());
template <class T, class U>
auto operator!=(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() != rhs.get()))
-> decltype(lhs.get() != rhs.get());
template <class T, class U>
auto operator<(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() < rhs.get()))
-> decltype(lhs.get() < rhs.get());
template <class T, class U>
auto operator<=(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() <= rhs.get()))
-> decltype(lhs.get() <= rhs.get());
template <class T, class U>
auto operator>(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() > rhs.get()))
-> decltype(lhs.get() > rhs.get());
template <class T, class U>
auto operator>=(const not_null<T>& lhs,
const not_null<U>& rhs) noexcept(noexcept(lhs.get() >= rhs.get()))
-> decltype(lhs.get() >= rhs.get());
Comparison of pointers that are convertible to each other.
template <class T>
std::ostream& operator<<(std::ostream& os, const not_null<T>& val);
Performs stream output on a not_null
pointer, invoking os << val.get()
. This function is only available when GSL_NO_IOSTREAMS
is not defined.
template <class T, class U>
std::ptrdiff_t operator-(const not_null<T>&, const not_null<U>&) = delete;
template <class T>
not_null<T> operator-(const not_null<T>&, std::ptrdiff_t) = delete;
template <class T>
not_null<T> operator+(const not_null<T>&, std::ptrdiff_t) = delete;
template <class T>
not_null<T> operator+(std::ptrdiff_t, const not_null<T>&) = delete;
Addition and subtraction are explicitly deleted. Pointers point to single objects (I.13: Do not pass an array as a single pointer), so don't allow these operators.
template <class T>
struct std::hash<gsl::not_null<T>> { ... };
Specialization of std::hash
for gsl::not_null
.
strict_not_null
is the same as not_null
except that the constructors are explicit
.
The free function that deduces the target type from the type of the argument and creates a gsl::strict_not_null
object is gsl::make_strict_not_null
.
This header file exports the class gsl::span
, a bounds-checked implementation of std::span
.
template <class ElementType, std::size_t Extent>
class span;
gsl::span
is a view over memory. It does not own the memory and is only a way to access contiguous sequences of objects.
The extent can be either a fixed size or gsl::dynamic_extent
.
The gsl::span
is based on the standardized version of std::span
which was added to C++20. Originally, the plan was to
deprecate gsl::span
when std::span
finished standardization, however that plan changed when the runtime bounds checking
was removed from std::span
's design.
The only difference between gsl::span
and std::span
is that gsl::span
strictly enforces runtime bounds checking.
Any violations of the bounds check results in termination of the program.
Like gsl::span
, gsl::span
's iterators also differ from std::span
's iterator in that all access operations are bounds checked.
- you want to guarantee bounds safety in your project.
- All data accessing operations use bounds checking to ensure you are only accessing valid memory.
- your project uses C++14 or C++17.
std::span
is not available as it was not introduced into the STL until C++20.
- your project is C++20 and you need the performance offered by
std::span
.
using element_type = ElementType;
using value_type = std::remove_cv_t<ElementType>;
using size_type = std::size_t;
using pointer = element_type*;
using const_pointer = const element_type*;
using reference = element_type&;
using const_reference = const element_type&;
using difference_type = std::ptrdiff_t;
using iterator = details::span_iterator<ElementType>;
using reverse_iterator = std::reverse_iterator<iterator>;
constexpr span() noexcept;
Constructs an empty span
. This constructor is only available if Extent
is 0 or gsl::dynamic_extent
.
span::data()
will return nullptr
.
constexpr explicit(Extent != gsl::dynamic_extent) span(pointer ptr, size_type count) noexcept;
Constructs a span
from a pointer and a size. If Extent
is not gsl::dynamic_extent
,
then the constructor Expects
that count == Extent
.
constexpr explicit(Extent != gsl::dynamic_extent) span(pointer firstElem, pointer lastElem) noexcept;
Constructs a span
from a pointer to the begin and the end of the data. If Extent
is not gsl::dynamic_extent
,
then the constructor Expects
that lastElem - firstElem == Extent
.
template <std::size_t N>
constexpr span(element_type (&arr)[N]) noexcept;
Constructs a span
from a C style array. This overload is available if Extent ==
gsl::dynamic_extent
or N == Extent
.
template <class T, std::size_t N>
constexpr span(std::array<T, N>& arr) noexcept;
template <class T, std::size_t N>
constexpr span(const std::array<T, N>& arr) noexcept;
Constructs a span
from a std::array
. These overloads are available if Extent ==
gsl::dynamic_extent
or N == Extent
, and if the array can be interpreted as a ElementType
array.
template <class Container>
constexpr explicit(Extent != gsl::dynamic_extent) span(Container& cont) noexcept;
template <class Container>
constexpr explicit(Extent != gsl::dynamic_extent) span(const Container& cont) noexcept;
Constructs a span
from a container. These overloads are available if Extent ==
gsl::dynamic_extent
or N == Extent
, and if the container can be interpreted as a contiguous ElementType
array.
constexpr span(const span& other) noexcept = default;
Copy constructor.
template <class OtherElementType, std::size_t OtherExtent>
explicit(Extent != gsl::dynamic_extent && OtherExtent == dynamic_extent)
constexpr span(const span<OtherElementType, OtherExtent>& other) noexcept;
Constructs a span
from another span
. This constructor is available if OtherExtent == Extent || Extent ==
gsl::dynamic_extent
|| OtherExtent ==
gsl::dynamic_extent
and if ElementType
and OtherElementType
are compatible.
If Extent !=
gsl::dynamic_extent
and OtherExtent ==
gsl::dynamic_extent
,
then the constructor Expects
that other.size() == Extent
.
constexpr span& operator=(const span& other) noexcept = default;
Copy assignment
template <std::size_t Count>
constexpr span<element_type, Count> first() const noexcept;
constexpr span<element_type, dynamic_extent> first(size_type count) const noexcept;
template <std::size_t Count>
constexpr span<element_type, Count> last() const noexcept;
constexpr span<element_type, dynamic_extent> last(size_type count) const noexcept;
Return a subspan of the first/last Count
elements. Expects
that Count
does not exceed the span
's size.
template <std::size_t offset, std::size_t count = dynamic_extent>
constexpr auto subspan() const noexcept;
constexpr span<element_type, dynamic_extent>
subspan(size_type offset, size_type count = dynamic_extent) const noexcept;
Return a subspan starting at offset
and having size count
. Expects
that offset
does not exceed the span
's size,
and that offset ==
gsl::dynamic_extent
or offset + count
does not exceed the span
's size.
If count
is gsl::dynamic_extent
, the number of elements in the subspan is size() - offset
.
constexpr size_type size() const noexcept;
constexpr size_type size_bytes() const noexcept;
Returns the size respective the size in bytes of the span
.
constexpr bool empty() const noexcept;
Is the span
empty?
constexpr reference operator[](size_type idx) const noexcept;
Returns a reference to the element at the given index. Expects
that idx
is less than the span
's size.
constexpr reference front() const noexcept;
constexpr reference back() const noexcept;
Returns a reference to the first/last element in the span
. Expects
that the span
is not empty.
constexpr pointer data() const noexcept;
Returns a pointer to the beginning of the contained data.
constexpr iterator begin() const noexcept;
constexpr iterator end() const noexcept;
constexpr reverse_iterator rbegin() const noexcept;
constexpr reverse_iterator rend() const noexcept;
Returns an iterator to the first/last normal/reverse iterator.
template <class Type, std::size_t Extent>
span(Type (&)[Extent]) -> span<Type, Extent>;
template <class Type, std::size_t Size>
span(std::array<Type, Size>&) -> span<Type, Size>;
template <class Type, std::size_t Size>
span(const std::array<Type, Size>&) -> span<const Type, Size>;
template <class Container,
class Element = std::remove_pointer_t<decltype(std::declval<Container&>().data())>>
span(Container&) -> span<Element>;
template <class Container,
class Element = std::remove_pointer_t<decltype(std::declval<const Container&>().data())>>
span(const Container&) -> span<Element>;
Deduction guides.
template <class ElementType, std::size_t Extent>
span<const byte, details::calculate_byte_size<ElementType, Extent>::value>
as_bytes(span<ElementType, Extent> s) noexcept;
template <class ElementType, std::size_t Extent>
span<byte, details::calculate_byte_size<ElementType, Extent>::value>
as_writable_bytes(span<ElementType, Extent> s) noexcept;
Converts a span
into a span
of byte
s.
as_writable_bytes
will only be available for non-const ElementType
s.
This file is a companion for and included by <gsl/span>
, and should not be used on its own. It contains useful features that aren't part of the std::span
API as found inside the STL <span>
header (with the exception of gsl::dynamic_extent
, which is included here due to implementation constraints).
gsl::dynamic_extent
gsl::span
gsl::span
comparison operatorsgsl::make_span
gsl::at
gsl::ssize
gsl::span
iterator functions
Defines the extent value to be used by all gsl::span
with dynamic extent.
Note: std::dynamic_extent
is exposed by the STL <span>
header and so ideally gsl::dynamic_extent
would be under <gsl/span>
, but to avoid cyclic dependency issues it is under <span_ext>
instead.
template <class ElementType, std::size_t Extent = dynamic_extent>
class span;
Forward declaration of gsl::span
.
template <class ElementType, std::size_t FirstExtent, std::size_t SecondExtent>
constexpr bool operator==(span<ElementType, FirstExtent> l, span<ElementType, SecondExtent> r);
template <class ElementType, std::size_t FirstExtent, std::size_t SecondExtent>
constexpr bool operator!=(span<ElementType, FirstExtent> l, span<ElementType, SecondExtent> r);
template <class ElementType, std::size_t Extent>
constexpr bool operator<(span<ElementType, Extent> l, span<ElementType, Extent> r);
template <class ElementType, std::size_t Extent>
constexpr bool operator<=(span<ElementType, Extent> l, span<ElementType, Extent> r);
template <class ElementType, std::size_t Extent>
constexpr bool operator>(span<ElementType, Extent> l, span<ElementType, Extent> r);
template <class ElementType, std::size_t Extent>
constexpr bool operator>=(span<ElementType, Extent> l, span<ElementType, Extent> r);
The comparison operators for two span
s lexicographically compare the elements in the span
s.
template <class ElementType>
constexpr span<ElementType> make_span(ElementType* ptr, typename span<ElementType>::size_type count);
template <class ElementType>
constexpr span<ElementType> make_span(ElementType* firstElem, ElementType* lastElem);
template <class ElementType, std::size_t N>
constexpr span<ElementType, N> make_span(ElementType (&arr)[N]) noexcept;
template <class Container>
constexpr span<typename Container::value_type> make_span(Container& cont);
template <class Container>
constexpr span<const typename Container::value_type> make_span(const Container& cont);
template <class Ptr>
constexpr span<typename Ptr::element_type> make_span(Ptr& cont, std::size_t count);
template <class Ptr>
constexpr span<typename Ptr::element_type> make_span(Ptr& cont);
Utility function for creating a span
with gsl::dynamic_extent
from
- pointer and length,
- pointer to start and pointer to end,
- a C style array, or
- a container.
template <class ElementType, std::size_t Extent>
constexpr ElementType& at(span<ElementType, Extent> s, index i);
The function gsl::at
offers a safe way to access data with index bounds checking.
This is the specialization of gsl::at
for span
. It returns a reference to the i
th element and
Expects
that the provided index is within the bounds of the span
.
Note: gsl::at
supports indexes up to PTRDIFF_MAX
.
template <class ElementType, std::size_t Extent>
constexpr std::ptrdiff_t ssize(const span<ElementType, Extent>& s) noexcept;
Return the size of a span
as a ptrdiff_t
.
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::iterator
begin(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent = dynamic_extent>
constexpr typename span<ElementType, Extent>::iterator
end(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::reverse_iterator
rbegin(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::reverse_iterator
rend(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::iterator
cbegin(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent = dynamic_extent>
constexpr typename span<ElementType, Extent>::iterator
cend(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::reverse_iterator
crbegin(const span<ElementType, Extent>& s) noexcept;
template <class ElementType, std::size_t Extent>
constexpr typename span<ElementType, Extent>::reverse_iterator
crend(const span<ElementType, Extent>& s) noexcept;
Free functions for getting a non-const/const begin/end normal/reverse iterator for a span
.
This header exports a family of *zstring
types.
A gsl::XXzstring<T>
is a typedef to T
. It adds no checks whatsoever, it is just for having a syntax to describe
that a pointer points to a zero terminated C style string. This helps static code analysis, and it helps human readers.
basic_zstring
is a pointer to a C-string (zero-terminated array) with a templated char type. Used to implement the rest of the *zstring
family.
zstring
is a zero terminated char
string.
czstring
is a const zero terminated char
string.
wzstring
is a zero terminated wchar_t
string.
cwzstring
is a const zero terminated wchar_t
string.
u16zstring
is a zero terminated char16_t
string.
cu16zstring
is a const zero terminated char16_t
string.
u32zstring
is a zero terminated char32_t
string.
cu32zstring
is a const zero terminated char32_t
string.
See GSL.view and SL.str.3: Use zstring or czstring to refer to a C-style, zero-terminated, sequence of characters.
This header contains utility functions and classes. This header works without exceptions being available. The parts that require exceptions being available are in their own header file narrow.
An alias to std::ptrdiff_t
. It serves as the index type for all container indexes/subscripts/sizes.
gsl::narrow_cast<T>(x)
is a named cast that is identical to a static_cast<T>(x)
. It exists to make clear to static code analysis tools and to human readers that a lossy conversion is acceptable.
Note: compare the throwing version gsl::narrow
in header narrow.
See ES.46: Avoid lossy (narrowing, truncating) arithmetic conversions and ES.49: If you must use a cast, use a named cast
template <class F>
class final_action { ... };
final_action
allows you to ensure something gets run at the end of a scope.
See E.19: Use a final_action object to express cleanup if no suitable resource handle is available
explicit final_action(const F& ff) noexcept;
explicit final_action(F&& ff) noexcept;
Construct an object with the action to invoke in the destructor.
~final_action() noexcept;
The destructor will call the action that was passed in the constructor.
final_action(final_action&& other) noexcept;
final_action(const final_action&) = delete;
void operator=(const final_action&) = delete;
void operator=(final_action&&) = delete;
Move construction is allowed. Copy construction is deleted. Copy and move assignment are also explicitly deleted.
template <class F>
auto finally(F&& f) noexcept;
Creates a gsl::final_action
object, deducing the template argument type from the type of the argument.
The function gsl::at
offers a safe way to access data with index bounds checking.
Note: gsl::at
supports indexes up to PTRDIFF_MAX
.
See ES.42: Keep use of pointers simple and straightforward
template <class T, std::size_t N>
constexpr T& at(T (&arr)[N], const index i);
This overload returns a reference to the i
s element of a C style array arr
. It Expects
that the provided index is within the bounds of the array.
template <class Cont>
constexpr auto at(Cont& cont, const index i) -> decltype(cont[cont.size()]);
This overload returns a reference to the i
s element of the container cont
. It Expects
that the provided index is within the bounds of the array.
template <class T>
constexpr T at(const std::initializer_list<T> cont, const index i);
This overload returns a reference to the i
s element of the initializer list cont
. It Expects
that the provided index is within the bounds of the array.
template <class T, std::size_t extent = std::dynamic_extent>
constexpr auto at(std::span<T, extent> sp, const index i) -> decltype(sp[sp.size()]);
This overload returns a reference to the i
s element of the std::span
sp
. It Expects
that the provided index is within the bounds of the array.