Simple device (fancy) pointer implementationTrying to find a good design for reading in values of different types from a fileUnique pointer implementationHeap implementation using pointerSimple shared pointerArray-like container for uints shorter than 8 bits (Rev 1)Alternate “weak” pointer implementationPreferred implementation of `Array<T>::operator=(const Array<T> & rhs)`C++ maybe pointer type implementationAttempt at Smart Pointer ImplementationC++: Smart Pointer Implementation
How to add power-LED to my small amplifier?
How to make payment on the internet without leaving a money trail?
Why don't electron-positron collisions release infinite energy?
What typically incentivizes a professor to change jobs to a lower ranking university?
New order #4: World
Is there a minimum number of transactions in a block?
Why doesn't Newton's third law mean a person bounces back to where they started when they hit the ground?
How does one intimidate enemies without having the capacity for violence?
Possibly bubble sort algorithm
Concept of linear mappings are confusing me
Why has Russell's definition of numbers using equivalence classes been finally abandoned? ( If it has actually been abandoned).
Should I join office cleaning event for free?
How much RAM could one put in a typical 80386 setup?
DOS, create pipe for stdin/stdout of command.com(or 4dos.com) in C or Batch?
What defenses are there against being summoned by the Gate spell?
I’m planning on buying a laser printer but concerned about the life cycle of toner in the machine
If Manufacturer spice model and Datasheet give different values which should I use?
Patience, young "Padovan"
How to type dʒ symbol (IPA) on Mac?
Prevent a directory in /tmp from being deleted
How can bays and straits be determined in a procedurally generated map?
Why was the small council so happy for Tyrion to become the Master of Coin?
I probably found a bug with the sudo apt install function
declaring a variable twice in IIFE
Simple device (fancy) pointer implementation
Trying to find a good design for reading in values of different types from a fileUnique pointer implementationHeap implementation using pointerSimple shared pointerArray-like container for uints shorter than 8 bits (Rev 1)Alternate “weak” pointer implementationPreferred implementation of `Array<T>::operator=(const Array<T> & rhs)`C++ maybe pointer type implementationAttempt at Smart Pointer ImplementationC++: Smart Pointer Implementation
.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty margin-bottom:0;
$begingroup$
device_raw_ptr
is a simple fancy pointer. It essentially wrap pointers to GPU memory. It's sole purpose is to separate out host pointers from device pointers, i.e. they should not be inter-convertible and must not be dereferenced. At the same time, they should be zero-cost (with respect to raw host pointers) and be maximally compatible with regular pointers.
Helper class:
template <class T>
struct equality_operators
/*
** The deriving class must implement the following:
** friend bool operator==(const T&, const T&);
*/
friend bool operator!=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs == rhs);
;
template <class T>
struct less_than_operators
/*
** The deriving class must implement the following:
** friend bool operator<(const T&, const T&);
*/
friend bool operator>(const T& lhs, const T& rhs) return rhs < lhs;
friend bool operator<=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs > rhs);
friend bool operator>=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs < rhs);
;
template <class T>
struct relational_operators : equality_operators<T>, less_than_operators<T> ;
device_raw_ptr
implementation:
template <class T>
class device_raw_ptr : public relational_operators<T>
static_assert(std::is_standard_layout<T>::value, "T must satisfy StandardLayoutType");
public:
using element_type = std::remove_extent_t<T>;
constexpr device_raw_ptr() noexcept = default;
constexpr device_raw_ptr(std::nullptr_t) noexcept : ptr nullptr
constexpr device_raw_ptr(const device_raw_ptr& other) noexcept = default;
explicit device_raw_ptr(element_type* ptr_) noexcept : ptr ptr_
device_raw_ptr(device_raw_ptr&& other) noexcept : ptr other.ptr other.reset();
device_raw_ptr& operator=(const device_raw_ptr& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
device_raw_ptr& operator=(device_raw_ptr&& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
void reset() noexcept ptr = nullptr;
void reset(T* ptr_) noexcept ptr = ptr_;
element_type* get() noexcept return ptr; ;
const element_type* get() const noexcept return ptr;
friend void swap(device_raw_ptr& lhs, device_raw_ptr& rhs) noexcept
using std::swap;
std::swap(lhs.ptr, rhs.ptr);
explicit operator bool() const noexcept return static_cast<bool>(ptr);
device_raw_ptr& operator++() noexcept
++ptr;
return *this;
device_raw_ptr operator++(int) noexcept
device_raw_ptr tmp(*this);
ptr++;
return tmp;
device_raw_ptr& operator+=(std::ptrdiff_t offset) noexcept
ptr += offset;
return *this;
device_raw_ptr& operator-=(std::ptrdiff_t offset) noexcept
ptr -= offset;
return *this;
friend device_raw_ptr& operator+(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs += offset;
return lhs;
friend device_raw_ptr& operator-(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs -= offset;
return lhs;
/* required by relational_operators base class */
friend bool operator==(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr == rhs.ptr;
friend bool operator<(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr < rhs.ptr;
protected:
T *ptr;
;
template <class T, class U, class V>
std::basic_ostream<U, V>& operator<<(std::basic_ostream<U, V>& os, const device_raw_ptr<T>& other)
os << other.get() << " (device)";
return os;
I am also looking for suggestions on how to order different things inside a class.
Might it be better to initialize with nullptr
instead of the default constructor. This breaks compatibility but I think it might be worth considering the compiler would mostly optimize the assignment if it's immediately overwritten.
c++ c++11 pointers
New contributor
$endgroup$
add a comment |
$begingroup$
device_raw_ptr
is a simple fancy pointer. It essentially wrap pointers to GPU memory. It's sole purpose is to separate out host pointers from device pointers, i.e. they should not be inter-convertible and must not be dereferenced. At the same time, they should be zero-cost (with respect to raw host pointers) and be maximally compatible with regular pointers.
Helper class:
template <class T>
struct equality_operators
/*
** The deriving class must implement the following:
** friend bool operator==(const T&, const T&);
*/
friend bool operator!=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs == rhs);
;
template <class T>
struct less_than_operators
/*
** The deriving class must implement the following:
** friend bool operator<(const T&, const T&);
*/
friend bool operator>(const T& lhs, const T& rhs) return rhs < lhs;
friend bool operator<=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs > rhs);
friend bool operator>=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs < rhs);
;
template <class T>
struct relational_operators : equality_operators<T>, less_than_operators<T> ;
device_raw_ptr
implementation:
template <class T>
class device_raw_ptr : public relational_operators<T>
static_assert(std::is_standard_layout<T>::value, "T must satisfy StandardLayoutType");
public:
using element_type = std::remove_extent_t<T>;
constexpr device_raw_ptr() noexcept = default;
constexpr device_raw_ptr(std::nullptr_t) noexcept : ptr nullptr
constexpr device_raw_ptr(const device_raw_ptr& other) noexcept = default;
explicit device_raw_ptr(element_type* ptr_) noexcept : ptr ptr_
device_raw_ptr(device_raw_ptr&& other) noexcept : ptr other.ptr other.reset();
device_raw_ptr& operator=(const device_raw_ptr& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
device_raw_ptr& operator=(device_raw_ptr&& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
void reset() noexcept ptr = nullptr;
void reset(T* ptr_) noexcept ptr = ptr_;
element_type* get() noexcept return ptr; ;
const element_type* get() const noexcept return ptr;
friend void swap(device_raw_ptr& lhs, device_raw_ptr& rhs) noexcept
using std::swap;
std::swap(lhs.ptr, rhs.ptr);
explicit operator bool() const noexcept return static_cast<bool>(ptr);
device_raw_ptr& operator++() noexcept
++ptr;
return *this;
device_raw_ptr operator++(int) noexcept
device_raw_ptr tmp(*this);
ptr++;
return tmp;
device_raw_ptr& operator+=(std::ptrdiff_t offset) noexcept
ptr += offset;
return *this;
device_raw_ptr& operator-=(std::ptrdiff_t offset) noexcept
ptr -= offset;
return *this;
friend device_raw_ptr& operator+(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs += offset;
return lhs;
friend device_raw_ptr& operator-(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs -= offset;
return lhs;
/* required by relational_operators base class */
friend bool operator==(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr == rhs.ptr;
friend bool operator<(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr < rhs.ptr;
protected:
T *ptr;
;
template <class T, class U, class V>
std::basic_ostream<U, V>& operator<<(std::basic_ostream<U, V>& os, const device_raw_ptr<T>& other)
os << other.get() << " (device)";
return os;
I am also looking for suggestions on how to order different things inside a class.
Might it be better to initialize with nullptr
instead of the default constructor. This breaks compatibility but I think it might be worth considering the compiler would mostly optimize the assignment if it's immediately overwritten.
c++ c++11 pointers
New contributor
$endgroup$
$begingroup$
I just noticed that I am passing rvalues toswap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.
$endgroup$
– Yashas
9 hours ago
add a comment |
$begingroup$
device_raw_ptr
is a simple fancy pointer. It essentially wrap pointers to GPU memory. It's sole purpose is to separate out host pointers from device pointers, i.e. they should not be inter-convertible and must not be dereferenced. At the same time, they should be zero-cost (with respect to raw host pointers) and be maximally compatible with regular pointers.
Helper class:
template <class T>
struct equality_operators
/*
** The deriving class must implement the following:
** friend bool operator==(const T&, const T&);
*/
friend bool operator!=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs == rhs);
;
template <class T>
struct less_than_operators
/*
** The deriving class must implement the following:
** friend bool operator<(const T&, const T&);
*/
friend bool operator>(const T& lhs, const T& rhs) return rhs < lhs;
friend bool operator<=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs > rhs);
friend bool operator>=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs < rhs);
;
template <class T>
struct relational_operators : equality_operators<T>, less_than_operators<T> ;
device_raw_ptr
implementation:
template <class T>
class device_raw_ptr : public relational_operators<T>
static_assert(std::is_standard_layout<T>::value, "T must satisfy StandardLayoutType");
public:
using element_type = std::remove_extent_t<T>;
constexpr device_raw_ptr() noexcept = default;
constexpr device_raw_ptr(std::nullptr_t) noexcept : ptr nullptr
constexpr device_raw_ptr(const device_raw_ptr& other) noexcept = default;
explicit device_raw_ptr(element_type* ptr_) noexcept : ptr ptr_
device_raw_ptr(device_raw_ptr&& other) noexcept : ptr other.ptr other.reset();
device_raw_ptr& operator=(const device_raw_ptr& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
device_raw_ptr& operator=(device_raw_ptr&& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
void reset() noexcept ptr = nullptr;
void reset(T* ptr_) noexcept ptr = ptr_;
element_type* get() noexcept return ptr; ;
const element_type* get() const noexcept return ptr;
friend void swap(device_raw_ptr& lhs, device_raw_ptr& rhs) noexcept
using std::swap;
std::swap(lhs.ptr, rhs.ptr);
explicit operator bool() const noexcept return static_cast<bool>(ptr);
device_raw_ptr& operator++() noexcept
++ptr;
return *this;
device_raw_ptr operator++(int) noexcept
device_raw_ptr tmp(*this);
ptr++;
return tmp;
device_raw_ptr& operator+=(std::ptrdiff_t offset) noexcept
ptr += offset;
return *this;
device_raw_ptr& operator-=(std::ptrdiff_t offset) noexcept
ptr -= offset;
return *this;
friend device_raw_ptr& operator+(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs += offset;
return lhs;
friend device_raw_ptr& operator-(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs -= offset;
return lhs;
/* required by relational_operators base class */
friend bool operator==(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr == rhs.ptr;
friend bool operator<(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr < rhs.ptr;
protected:
T *ptr;
;
template <class T, class U, class V>
std::basic_ostream<U, V>& operator<<(std::basic_ostream<U, V>& os, const device_raw_ptr<T>& other)
os << other.get() << " (device)";
return os;
I am also looking for suggestions on how to order different things inside a class.
Might it be better to initialize with nullptr
instead of the default constructor. This breaks compatibility but I think it might be worth considering the compiler would mostly optimize the assignment if it's immediately overwritten.
c++ c++11 pointers
New contributor
$endgroup$
device_raw_ptr
is a simple fancy pointer. It essentially wrap pointers to GPU memory. It's sole purpose is to separate out host pointers from device pointers, i.e. they should not be inter-convertible and must not be dereferenced. At the same time, they should be zero-cost (with respect to raw host pointers) and be maximally compatible with regular pointers.
Helper class:
template <class T>
struct equality_operators
/*
** The deriving class must implement the following:
** friend bool operator==(const T&, const T&);
*/
friend bool operator!=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs == rhs);
;
template <class T>
struct less_than_operators
/*
** The deriving class must implement the following:
** friend bool operator<(const T&, const T&);
*/
friend bool operator>(const T& lhs, const T& rhs) return rhs < lhs;
friend bool operator<=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs > rhs);
friend bool operator>=(const T& lhs, const T& rhs) return !static_cast<bool>(lhs < rhs);
;
template <class T>
struct relational_operators : equality_operators<T>, less_than_operators<T> ;
device_raw_ptr
implementation:
template <class T>
class device_raw_ptr : public relational_operators<T>
static_assert(std::is_standard_layout<T>::value, "T must satisfy StandardLayoutType");
public:
using element_type = std::remove_extent_t<T>;
constexpr device_raw_ptr() noexcept = default;
constexpr device_raw_ptr(std::nullptr_t) noexcept : ptr nullptr
constexpr device_raw_ptr(const device_raw_ptr& other) noexcept = default;
explicit device_raw_ptr(element_type* ptr_) noexcept : ptr ptr_
device_raw_ptr(device_raw_ptr&& other) noexcept : ptr other.ptr other.reset();
device_raw_ptr& operator=(const device_raw_ptr& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
device_raw_ptr& operator=(device_raw_ptr&& other) noexcept
swap(device_raw_ptr(other), *this);
return *this;
void reset() noexcept ptr = nullptr;
void reset(T* ptr_) noexcept ptr = ptr_;
element_type* get() noexcept return ptr; ;
const element_type* get() const noexcept return ptr;
friend void swap(device_raw_ptr& lhs, device_raw_ptr& rhs) noexcept
using std::swap;
std::swap(lhs.ptr, rhs.ptr);
explicit operator bool() const noexcept return static_cast<bool>(ptr);
device_raw_ptr& operator++() noexcept
++ptr;
return *this;
device_raw_ptr operator++(int) noexcept
device_raw_ptr tmp(*this);
ptr++;
return tmp;
device_raw_ptr& operator+=(std::ptrdiff_t offset) noexcept
ptr += offset;
return *this;
device_raw_ptr& operator-=(std::ptrdiff_t offset) noexcept
ptr -= offset;
return *this;
friend device_raw_ptr& operator+(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs += offset;
return lhs;
friend device_raw_ptr& operator-(device_raw_ptr lhs, std::ptrdiff_t offset) noexcept
lhs -= offset;
return lhs;
/* required by relational_operators base class */
friend bool operator==(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr == rhs.ptr;
friend bool operator<(const device_raw_ptr& lhs, const device_raw_ptr& rhs) noexcept return lhs.ptr < rhs.ptr;
protected:
T *ptr;
;
template <class T, class U, class V>
std::basic_ostream<U, V>& operator<<(std::basic_ostream<U, V>& os, const device_raw_ptr<T>& other)
os << other.get() << " (device)";
return os;
I am also looking for suggestions on how to order different things inside a class.
Might it be better to initialize with nullptr
instead of the default constructor. This breaks compatibility but I think it might be worth considering the compiler would mostly optimize the assignment if it's immediately overwritten.
c++ c++11 pointers
c++ c++11 pointers
New contributor
New contributor
edited 9 hours ago
esote
3,01411039
3,01411039
New contributor
asked 10 hours ago
YashasYashas
1113
1113
New contributor
New contributor
$begingroup$
I just noticed that I am passing rvalues toswap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.
$endgroup$
– Yashas
9 hours ago
add a comment |
$begingroup$
I just noticed that I am passing rvalues toswap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.
$endgroup$
– Yashas
9 hours ago
$begingroup$
I just noticed that I am passing rvalues to
swap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.$endgroup$
– Yashas
9 hours ago
$begingroup$
I just noticed that I am passing rvalues to
swap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.$endgroup$
– Yashas
9 hours ago
add a comment |
2 Answers
2
active
oldest
votes
$begingroup$
Helper classes
There is no need to use static_cast<bool>
for your comparisons. The relational operators are already bool
values. (If they are not, that is a problem with the definition of the operator for the type T
.)
The standard <utility>
header provides definitions for operator!=
(from operator==
) and operator>
, operator<=
, and operator>=
(from operator<
). There is no need for you to define those four operators if you have the other two (equality and less-than).
Why do you have the relational_operators
struct at all? It shouldn't be necessary.
Implementation
The default constructor for device_raw_ptr
leaves the ptr
member uninitialized. Typically a class like this would initialize ptr
to nullptr
, and you wouldn't need the constructor that takes a std::nullptr_t
object.
The copy assignment operator should just be ptr = other.ptr
, since that is the only thing in your class. The way you have it is nonstandard behavior. You construct a temporary, then pass it as a non-const reference to swap
. This is not supported as part of the language, although some compilers (MSVC) support it as an extension. You're constructing a temporary, doing a swap, then destroying the temporary (a noop in this case). Similarly, the move assignment operator can be simplified to not use the temporary (ptr = other.ptr; other.reset();
, or use three statements with an assignment to a local to avoid problems if you move assign an object to itself).
operator bool
does not need a static_cast
. Perhaps an explicit ptr != nullptr
check, although a pointer will implicitly convert to a bool
.
$endgroup$
$begingroup$
I thought usingstd::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests usingboost:operators
. I wrote my own version instead.
$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leavesptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it tonullptr
. Thestd::nullptr_t
is a consequence of the aforementioned statement. I intended to have aconstexpr
constructor which setsptr
tonullptr
.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashasstd::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and= delete
the ones it doesn't want to support. But its up to you to decide how to implement them.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning anullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.
$endgroup$
– 1201ProgramAlarm
9 hours ago
add a comment |
$begingroup$
device_raw_ptr
is extremely cheap to copy, so remove all hints of move-semantics and use ofswap()
.Now you can remove the copy-constructor and copy-assignment, as there is no need to explicitly declare them. Especially making them user-defined must be avoided to keep them trivial.
Kudos on trying to use the approved two-step for
swap()
. Though you get a failing grade anyway because you bungled it by using a qualified call in the second part. Hopefully, C++20 will abolish that nonsense by introducing customization point objects.Yes, you should pass your
device_raw_ptr
by value if you have the choice, as it is a tiny trivial type. Still, refrain from returning a reference to such a temporary.There is no reason
operator-(device_raw_ptr, std::ptrdiff_t)
should not beconstexpr
. Aside from your implementation for some reason delegating tooperator-=
, which is not. Same foroperator+
which usesoperator+=
.Is there any reason you don't support subtracting a
device_raw_ptr
from another?I'm really puzzled why you make the only data-member
protected
instead ofprivate
.
$endgroup$
add a comment |
Your Answer
StackExchange.ifUsing("editor", function ()
return StackExchange.using("mathjaxEditing", function ()
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix)
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["\$", "\$"]]);
);
);
, "mathjax-editing");
StackExchange.ifUsing("editor", function ()
StackExchange.using("externalEditor", function ()
StackExchange.using("snippets", function ()
StackExchange.snippets.init();
);
);
, "code-snippets");
StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "196"
;
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function()
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled)
StackExchange.using("snippets", function()
createEditor();
);
else
createEditor();
);
function createEditor()
StackExchange.prepareEditor(
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: false,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: null,
bindNavPrevention: true,
postfix: "",
imageUploader:
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
,
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
);
);
Yashas is a new contributor. Be nice, and check out our Code of Conduct.
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fcodereview.stackexchange.com%2fquestions%2f217019%2fsimple-device-fancy-pointer-implementation%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
2 Answers
2
active
oldest
votes
2 Answers
2
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
Helper classes
There is no need to use static_cast<bool>
for your comparisons. The relational operators are already bool
values. (If they are not, that is a problem with the definition of the operator for the type T
.)
The standard <utility>
header provides definitions for operator!=
(from operator==
) and operator>
, operator<=
, and operator>=
(from operator<
). There is no need for you to define those four operators if you have the other two (equality and less-than).
Why do you have the relational_operators
struct at all? It shouldn't be necessary.
Implementation
The default constructor for device_raw_ptr
leaves the ptr
member uninitialized. Typically a class like this would initialize ptr
to nullptr
, and you wouldn't need the constructor that takes a std::nullptr_t
object.
The copy assignment operator should just be ptr = other.ptr
, since that is the only thing in your class. The way you have it is nonstandard behavior. You construct a temporary, then pass it as a non-const reference to swap
. This is not supported as part of the language, although some compilers (MSVC) support it as an extension. You're constructing a temporary, doing a swap, then destroying the temporary (a noop in this case). Similarly, the move assignment operator can be simplified to not use the temporary (ptr = other.ptr; other.reset();
, or use three statements with an assignment to a local to avoid problems if you move assign an object to itself).
operator bool
does not need a static_cast
. Perhaps an explicit ptr != nullptr
check, although a pointer will implicitly convert to a bool
.
$endgroup$
$begingroup$
I thought usingstd::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests usingboost:operators
. I wrote my own version instead.
$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leavesptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it tonullptr
. Thestd::nullptr_t
is a consequence of the aforementioned statement. I intended to have aconstexpr
constructor which setsptr
tonullptr
.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashasstd::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and= delete
the ones it doesn't want to support. But its up to you to decide how to implement them.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning anullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.
$endgroup$
– 1201ProgramAlarm
9 hours ago
add a comment |
$begingroup$
Helper classes
There is no need to use static_cast<bool>
for your comparisons. The relational operators are already bool
values. (If they are not, that is a problem with the definition of the operator for the type T
.)
The standard <utility>
header provides definitions for operator!=
(from operator==
) and operator>
, operator<=
, and operator>=
(from operator<
). There is no need for you to define those four operators if you have the other two (equality and less-than).
Why do you have the relational_operators
struct at all? It shouldn't be necessary.
Implementation
The default constructor for device_raw_ptr
leaves the ptr
member uninitialized. Typically a class like this would initialize ptr
to nullptr
, and you wouldn't need the constructor that takes a std::nullptr_t
object.
The copy assignment operator should just be ptr = other.ptr
, since that is the only thing in your class. The way you have it is nonstandard behavior. You construct a temporary, then pass it as a non-const reference to swap
. This is not supported as part of the language, although some compilers (MSVC) support it as an extension. You're constructing a temporary, doing a swap, then destroying the temporary (a noop in this case). Similarly, the move assignment operator can be simplified to not use the temporary (ptr = other.ptr; other.reset();
, or use three statements with an assignment to a local to avoid problems if you move assign an object to itself).
operator bool
does not need a static_cast
. Perhaps an explicit ptr != nullptr
check, although a pointer will implicitly convert to a bool
.
$endgroup$
$begingroup$
I thought usingstd::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests usingboost:operators
. I wrote my own version instead.
$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leavesptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it tonullptr
. Thestd::nullptr_t
is a consequence of the aforementioned statement. I intended to have aconstexpr
constructor which setsptr
tonullptr
.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashasstd::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and= delete
the ones it doesn't want to support. But its up to you to decide how to implement them.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning anullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.
$endgroup$
– 1201ProgramAlarm
9 hours ago
add a comment |
$begingroup$
Helper classes
There is no need to use static_cast<bool>
for your comparisons. The relational operators are already bool
values. (If they are not, that is a problem with the definition of the operator for the type T
.)
The standard <utility>
header provides definitions for operator!=
(from operator==
) and operator>
, operator<=
, and operator>=
(from operator<
). There is no need for you to define those four operators if you have the other two (equality and less-than).
Why do you have the relational_operators
struct at all? It shouldn't be necessary.
Implementation
The default constructor for device_raw_ptr
leaves the ptr
member uninitialized. Typically a class like this would initialize ptr
to nullptr
, and you wouldn't need the constructor that takes a std::nullptr_t
object.
The copy assignment operator should just be ptr = other.ptr
, since that is the only thing in your class. The way you have it is nonstandard behavior. You construct a temporary, then pass it as a non-const reference to swap
. This is not supported as part of the language, although some compilers (MSVC) support it as an extension. You're constructing a temporary, doing a swap, then destroying the temporary (a noop in this case). Similarly, the move assignment operator can be simplified to not use the temporary (ptr = other.ptr; other.reset();
, or use three statements with an assignment to a local to avoid problems if you move assign an object to itself).
operator bool
does not need a static_cast
. Perhaps an explicit ptr != nullptr
check, although a pointer will implicitly convert to a bool
.
$endgroup$
Helper classes
There is no need to use static_cast<bool>
for your comparisons. The relational operators are already bool
values. (If they are not, that is a problem with the definition of the operator for the type T
.)
The standard <utility>
header provides definitions for operator!=
(from operator==
) and operator>
, operator<=
, and operator>=
(from operator<
). There is no need for you to define those four operators if you have the other two (equality and less-than).
Why do you have the relational_operators
struct at all? It shouldn't be necessary.
Implementation
The default constructor for device_raw_ptr
leaves the ptr
member uninitialized. Typically a class like this would initialize ptr
to nullptr
, and you wouldn't need the constructor that takes a std::nullptr_t
object.
The copy assignment operator should just be ptr = other.ptr
, since that is the only thing in your class. The way you have it is nonstandard behavior. You construct a temporary, then pass it as a non-const reference to swap
. This is not supported as part of the language, although some compilers (MSVC) support it as an extension. You're constructing a temporary, doing a swap, then destroying the temporary (a noop in this case). Similarly, the move assignment operator can be simplified to not use the temporary (ptr = other.ptr; other.reset();
, or use three statements with an assignment to a local to avoid problems if you move assign an object to itself).
operator bool
does not need a static_cast
. Perhaps an explicit ptr != nullptr
check, although a pointer will implicitly convert to a bool
.
answered 9 hours ago
1201ProgramAlarm1201ProgramAlarm
3,7032925
3,7032925
$begingroup$
I thought usingstd::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests usingboost:operators
. I wrote my own version instead.
$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leavesptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it tonullptr
. Thestd::nullptr_t
is a consequence of the aforementioned statement. I intended to have aconstexpr
constructor which setsptr
tonullptr
.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashasstd::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and= delete
the ones it doesn't want to support. But its up to you to decide how to implement them.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning anullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.
$endgroup$
– 1201ProgramAlarm
9 hours ago
add a comment |
$begingroup$
I thought usingstd::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests usingboost:operators
. I wrote my own version instead.
$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leavesptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it tonullptr
. Thestd::nullptr_t
is a consequence of the aforementioned statement. I intended to have aconstexpr
constructor which setsptr
tonullptr
.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashasstd::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and= delete
the ones it doesn't want to support. But its up to you to decide how to implement them.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning anullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.
$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
I thought using
std::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests using boost:operators
. I wrote my own version instead.$endgroup$
– Yashas
9 hours ago
$begingroup$
I thought using
std::rel_ops
is generally frowned upon. stackoverflow.com/questions/6225375/idiomatic-use-of-stdrel-ops suggests using boost:operators
. I wrote my own version instead.$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leaves
ptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it to nullptr
. The std::nullptr_t
is a consequence of the aforementioned statement. I intended to have a constexpr
constructor which sets ptr
to nullptr
.$endgroup$
– Yashas
9 hours ago
$begingroup$
The default constructor leaves
ptr
unitialized because that's how raw pointers behave (initialized with garbage?). But I am considering the option of initializing it to nullptr
. The std::nullptr_t
is a consequence of the aforementioned statement. I intended to have a constexpr
constructor which sets ptr
to nullptr
.$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
For the abuse of swap, I made started (codereview.stackexchange.com/questions/217019/…) using temporaries right after I posted the question. It was an attempt to reuse the constructors to perform move/copy but I think it was overkill for such a simple class.
$endgroup$
– Yashas
9 hours ago
$begingroup$
@Yashas
std::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and = delete
the ones it doesn't want to support. But its up to you to decide how to implement them.$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas
std::rel_ops
is there to avoid having duplicate definitions of those relational operators. If a class does not want some of them (as mentioned in your linked question), then it should define all of them, and = delete
the ones it doesn't want to support. But its up to you to decide how to implement them.$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning a
nullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.$endgroup$
– 1201ProgramAlarm
9 hours ago
$begingroup$
@Yashas leaving raw pointer uninitialized is, generally, a bad idea. The cost of assigning a
nullptr
someplace where it isn't strictly necessary is outweighed by the predictability and consistent (mis)behavior the NULL value will give you.$endgroup$
– 1201ProgramAlarm
9 hours ago
add a comment |
$begingroup$
device_raw_ptr
is extremely cheap to copy, so remove all hints of move-semantics and use ofswap()
.Now you can remove the copy-constructor and copy-assignment, as there is no need to explicitly declare them. Especially making them user-defined must be avoided to keep them trivial.
Kudos on trying to use the approved two-step for
swap()
. Though you get a failing grade anyway because you bungled it by using a qualified call in the second part. Hopefully, C++20 will abolish that nonsense by introducing customization point objects.Yes, you should pass your
device_raw_ptr
by value if you have the choice, as it is a tiny trivial type. Still, refrain from returning a reference to such a temporary.There is no reason
operator-(device_raw_ptr, std::ptrdiff_t)
should not beconstexpr
. Aside from your implementation for some reason delegating tooperator-=
, which is not. Same foroperator+
which usesoperator+=
.Is there any reason you don't support subtracting a
device_raw_ptr
from another?I'm really puzzled why you make the only data-member
protected
instead ofprivate
.
$endgroup$
add a comment |
$begingroup$
device_raw_ptr
is extremely cheap to copy, so remove all hints of move-semantics and use ofswap()
.Now you can remove the copy-constructor and copy-assignment, as there is no need to explicitly declare them. Especially making them user-defined must be avoided to keep them trivial.
Kudos on trying to use the approved two-step for
swap()
. Though you get a failing grade anyway because you bungled it by using a qualified call in the second part. Hopefully, C++20 will abolish that nonsense by introducing customization point objects.Yes, you should pass your
device_raw_ptr
by value if you have the choice, as it is a tiny trivial type. Still, refrain from returning a reference to such a temporary.There is no reason
operator-(device_raw_ptr, std::ptrdiff_t)
should not beconstexpr
. Aside from your implementation for some reason delegating tooperator-=
, which is not. Same foroperator+
which usesoperator+=
.Is there any reason you don't support subtracting a
device_raw_ptr
from another?I'm really puzzled why you make the only data-member
protected
instead ofprivate
.
$endgroup$
add a comment |
$begingroup$
device_raw_ptr
is extremely cheap to copy, so remove all hints of move-semantics and use ofswap()
.Now you can remove the copy-constructor and copy-assignment, as there is no need to explicitly declare them. Especially making them user-defined must be avoided to keep them trivial.
Kudos on trying to use the approved two-step for
swap()
. Though you get a failing grade anyway because you bungled it by using a qualified call in the second part. Hopefully, C++20 will abolish that nonsense by introducing customization point objects.Yes, you should pass your
device_raw_ptr
by value if you have the choice, as it is a tiny trivial type. Still, refrain from returning a reference to such a temporary.There is no reason
operator-(device_raw_ptr, std::ptrdiff_t)
should not beconstexpr
. Aside from your implementation for some reason delegating tooperator-=
, which is not. Same foroperator+
which usesoperator+=
.Is there any reason you don't support subtracting a
device_raw_ptr
from another?I'm really puzzled why you make the only data-member
protected
instead ofprivate
.
$endgroup$
device_raw_ptr
is extremely cheap to copy, so remove all hints of move-semantics and use ofswap()
.Now you can remove the copy-constructor and copy-assignment, as there is no need to explicitly declare them. Especially making them user-defined must be avoided to keep them trivial.
Kudos on trying to use the approved two-step for
swap()
. Though you get a failing grade anyway because you bungled it by using a qualified call in the second part. Hopefully, C++20 will abolish that nonsense by introducing customization point objects.Yes, you should pass your
device_raw_ptr
by value if you have the choice, as it is a tiny trivial type. Still, refrain from returning a reference to such a temporary.There is no reason
operator-(device_raw_ptr, std::ptrdiff_t)
should not beconstexpr
. Aside from your implementation for some reason delegating tooperator-=
, which is not. Same foroperator+
which usesoperator+=
.Is there any reason you don't support subtracting a
device_raw_ptr
from another?I'm really puzzled why you make the only data-member
protected
instead ofprivate
.
answered 6 hours ago
DeduplicatorDeduplicator
11.8k1950
11.8k1950
add a comment |
add a comment |
Yashas is a new contributor. Be nice, and check out our Code of Conduct.
Yashas is a new contributor. Be nice, and check out our Code of Conduct.
Yashas is a new contributor. Be nice, and check out our Code of Conduct.
Yashas is a new contributor. Be nice, and check out our Code of Conduct.
Thanks for contributing an answer to Code Review Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fcodereview.stackexchange.com%2fquestions%2f217019%2fsimple-device-fancy-pointer-implementation%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
$begingroup$
I just noticed that I am passing rvalues to
swap
which accepts non-const lvalue as arguments. You can find it in the copy & move constructor.$endgroup$
– Yashas
9 hours ago