Date: Wed, 5 Mar 2025 15:14:08 +0100
> On 5 Mar 2025, at 14:44, Sebastian Wittmeier via Std-Proposals <std-proposals_at_[hidden]> wrote:
>
> Yes, I saw those examples.
> Error states, FP round-off errors and so on are details, which could be prohibited or better defined for pure functions.
My thinking is that floating-point round-off errors should not be considered an impurity, because one should write code so that it does not matter. Similarly, for error codes.
> If we fully embrace pure functions, we would provide the whole state as a read-only value parameter.
> The possibly new modified state would be output together with the result.
C++17 copy elision is strictly a side-effect, but is done to emulate a return value, so is therefore considered pure.
> A push_back on a vector would not modfiy the vector in-place, but output the appended vector as a copy.
> Pure functions can improve the performance (by caching, by better parallelization, by pipelining computations),
Pure functions to enable the use of threads are important, but possibly there might be different kinds of purity, like for optimizing.
> but they also can worsen it with the copy operations.
If a type is semantically values, then using impure code may lead to hard-to-catch programming errors. One example of that is the use of reference counting.
>
> Yes, I saw those examples.
> Error states, FP round-off errors and so on are details, which could be prohibited or better defined for pure functions.
My thinking is that floating-point round-off errors should not be considered an impurity, because one should write code so that it does not matter. Similarly, for error codes.
> If we fully embrace pure functions, we would provide the whole state as a read-only value parameter.
> The possibly new modified state would be output together with the result.
C++17 copy elision is strictly a side-effect, but is done to emulate a return value, so is therefore considered pure.
> A push_back on a vector would not modfiy the vector in-place, but output the appended vector as a copy.
> Pure functions can improve the performance (by caching, by better parallelization, by pipelining computations),
Pure functions to enable the use of threads are important, but possibly there might be different kinds of purity, like for optimizing.
> but they also can worsen it with the copy operations.
If a type is semantically values, then using impure code may lead to hard-to-catch programming errors. One example of that is the use of reference counting.
Received on 2025-03-05 14:14:21