I'm beginning to suspect that there is no one right answer to it
That would be my thought too :-) I think which operations can be supported without invalidating the iterator depends very much on how the underlying data structure is implemented. Also the solution depends on your requirements: when do you want the iterators to be invalidated - do you really need every thread to always see the latest version of the data, or are some changes (you mention value updates) acceptable? Plus, there might be performance considerations (i.e. lock the whole hash vs. selective locking for selected operations)?
Anyway, I just thought I'd mention that I've found Java's java.util.concurrent classes fairly well thought out, just for example ConcurrentHashMap or CopyOnWriteArrayList, perhaps that could serve as a bit of inspiration...
In reply to Re: Design thoughts: iterator invalidation
by Anonymous Monk
in thread Design thoughts: iterator invalidation
by BrowserUk
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |