On Wed, 31 May 2023, 03:22 Sebastian Wittmeier via Std-Proposals, <std-proposals@lists.isocpp.org> wrote:

I am not sure you understood the point of most of the replies: The level of detail for thread synchronization is the wrong one, if you do it on container level.

 

Compare it to database transactions. Why should the user of a relational database deal with them, if you could have every SQL command implicitly span a transaction and be done with it?

 

Because you want and need to combine a number of commands into one transaction. And those are application-specific. Should the flight reservation be automatically canceled if the payment fails? Different tables are involved, but the C++ standard committee would not know beforehand.

 

C++ could be extended with ways to simplify creating those middle layers, but you cannot spare the programmer from writing the actual application in the end.

 

Just having a container, where every call is thread-safe by itself, is the solution, which is nearly never (without proof of that by statistical data) needed. So it probably does not warrant to have classes for it in the standard library. Especially to also not give a false sense of security that the code would be thread-safe.

 

It is simple to generate your own wrapper or access classes or functions, if you need that feature.


Or use std::experimental::synchronized_value<std::vector<T>>