Re: Concurrent Containers
"Scott Meyers" <NeverRead@aristeia.com>
Java's ConcurrentHashMap, for example, has a size method that returns only
an approximation of the number of elements in the map, because, as Brian
Goetz puts it in "Java Concurrency in Practice," "the result of size could
be out of date by the time it is computed."
Ah, Bingo!!!
My previous question on "define correct in the context" was supposed to
bring out something like this.
My normal definition of correct size() for a container to know *the* number.
That is NOT delivered by the concurrent one. It fairly says so, meeting what
you answered a little up. But IMO this does not really meet what
programmers would "want" or expect in a real situation. rendering the
interface either being ballast or a landmine for too many potential
situations.
What I'm really interested in is the availability of
already-implemented-and-proven commonly-useful concurrency-friendly data
structure types for C++.
For this, as C++ gains native threading and move semantics hopefully soon I
expect such libraries/tools will finally emerge in the next few years.