Date: Sun, 5 Jan 2020 12:16:01 +0300
This is exactly what C++20 "spaceship operator" does.
сб, 4 янв. 2020 г. в 21:04, Walt Karas via Std-Proposals <
std-proposals_at_[hidden]>:
> Message: 1
> Date: Thu, 26 Dec 2019 13:31:45 -0500
> From: Edward Karak <edkharakh_at_[hidden]>
> To: std-proposals_at_[hidden]
> Subject: [std-proposals] Auto-generated comparison operators in class
> Message-ID:
> <CAMsy88+7i2Lx2HcWsxuUCezqPYu67JQ-pNp=TfX39DkxyBn6ag_at_[hidden]>
> Content-Type: text/plain; charset="utf-8"
>
> We often want to compare class instances to each other. For example, two
> date class instances. However, we have to tediously write out the
> implementations for operator>, operator<, operator==, operator>=, etc.
> which all have the same exact comparison logic, but simply a different
> return value. This is tedious and distracts the programmer.
>
> Instead, there should be a mechanism to write a single private comparison
> member function that behaves like the comparator in qsort, returning 1, 0
> or -1. Then the compiler would automatically generate all the relational
> operator overloads. Here's an example:
>
> class date {
> int yr, mo, day;
> /* new keyword: generate. Signals to the compiler that this method will
> provide the logic of the comparisons. */
> friend generate int cmp(const date &lhs, const date &rhs);
> public:
> /* constructors, etc. as usual */
> };
>
> This would require the use of a new keyword, which I've called generate
> (maybe
> a better name can be found that would minimize collision). This keyword is
> used so the compiler knows that it needs to auto-generate the relational
> operators for a class, and which function to call for implementing the
> comparison logic itself.
>
> The operators that the compiler will auto generate will look like this:
>
> friend bool operator==(const date &lhs, const date &rhs)
> {
> return cmp(lhs, rhs) == 0;
> }
>
> and so on and so forth for each relational operator.
>
> If, for some reason, a programmer wants to generate all relational
> operators *except* certain ones, then they can use the =delete notation. So
> for the date class above, let's say we want to auto-generate all relational
> operators except operator!=. Then the class would look like this:
>
> class date {
> int yr, mo, day;
> /* new keyword: generate. Signals to the compiler that this method will
> provide the logic of the comparisons. */
> friend generate int cmp(const date &lhs, const date &rhs);
> public:
> /* constructors, etc. as usual */
> friend bool operator!=(const date &lhs, const date &rhs) = delete;
> }
>
> The only objections I can anticipate to this proposal is that this feature
> would (1) significantly increase compilation times, as the compiler must
> generate new code; and (2) that a new keyword would be added, which may
> cause name clashes.
>
> However, the existence of templates, which auto-generates far more code
> than this would, refutes point (1). And point (2) can also be mitigated
> because the Standard can alternatively require a certain name for the
> comparison method, such as cmp, similar to how main is required to be the
> program entry point. This would eliminate the need to invent a new keyword.
> -------------- next part --------------
> HTML attachment scrubbed and removed
>
> ------------------------------
>
> I took a stab at doing something like this using the current base language:
>
>
> https://www.linkedin.com/pulse/easy-c-comparison-operators-udts-walter-karas/
>
>
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>
сб, 4 янв. 2020 г. в 21:04, Walt Karas via Std-Proposals <
std-proposals_at_[hidden]>:
> Message: 1
> Date: Thu, 26 Dec 2019 13:31:45 -0500
> From: Edward Karak <edkharakh_at_[hidden]>
> To: std-proposals_at_[hidden]
> Subject: [std-proposals] Auto-generated comparison operators in class
> Message-ID:
> <CAMsy88+7i2Lx2HcWsxuUCezqPYu67JQ-pNp=TfX39DkxyBn6ag_at_[hidden]>
> Content-Type: text/plain; charset="utf-8"
>
> We often want to compare class instances to each other. For example, two
> date class instances. However, we have to tediously write out the
> implementations for operator>, operator<, operator==, operator>=, etc.
> which all have the same exact comparison logic, but simply a different
> return value. This is tedious and distracts the programmer.
>
> Instead, there should be a mechanism to write a single private comparison
> member function that behaves like the comparator in qsort, returning 1, 0
> or -1. Then the compiler would automatically generate all the relational
> operator overloads. Here's an example:
>
> class date {
> int yr, mo, day;
> /* new keyword: generate. Signals to the compiler that this method will
> provide the logic of the comparisons. */
> friend generate int cmp(const date &lhs, const date &rhs);
> public:
> /* constructors, etc. as usual */
> };
>
> This would require the use of a new keyword, which I've called generate
> (maybe
> a better name can be found that would minimize collision). This keyword is
> used so the compiler knows that it needs to auto-generate the relational
> operators for a class, and which function to call for implementing the
> comparison logic itself.
>
> The operators that the compiler will auto generate will look like this:
>
> friend bool operator==(const date &lhs, const date &rhs)
> {
> return cmp(lhs, rhs) == 0;
> }
>
> and so on and so forth for each relational operator.
>
> If, for some reason, a programmer wants to generate all relational
> operators *except* certain ones, then they can use the =delete notation. So
> for the date class above, let's say we want to auto-generate all relational
> operators except operator!=. Then the class would look like this:
>
> class date {
> int yr, mo, day;
> /* new keyword: generate. Signals to the compiler that this method will
> provide the logic of the comparisons. */
> friend generate int cmp(const date &lhs, const date &rhs);
> public:
> /* constructors, etc. as usual */
> friend bool operator!=(const date &lhs, const date &rhs) = delete;
> }
>
> The only objections I can anticipate to this proposal is that this feature
> would (1) significantly increase compilation times, as the compiler must
> generate new code; and (2) that a new keyword would be added, which may
> cause name clashes.
>
> However, the existence of templates, which auto-generates far more code
> than this would, refutes point (1). And point (2) can also be mitigated
> because the Standard can alternatively require a certain name for the
> comparison method, such as cmp, similar to how main is required to be the
> program entry point. This would eliminate the need to invent a new keyword.
> -------------- next part --------------
> HTML attachment scrubbed and removed
>
> ------------------------------
>
> I took a stab at doing something like this using the current base language:
>
>
> https://www.linkedin.com/pulse/easy-c-comparison-operators-udts-walter-karas/
>
>
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>
Received on 2020-01-05 03:18:45