Walter is right about transitive readonly - here's the alternative
Jason House
jason.james.house at gmail.com
Tue Sep 18 18:00:02 PDT 2007
Janice Caron wrote:
> - - - - conclusion - - - -
>
> This is a viable, threadsafe alternative to both intransitive const
> and logical const.
I like your idea, but I have to disagree with this. The proposal itself
still needs some kind of const concept. How do you enforce something
which has a read lock is not written to?
I think a proposal that combines this and the latest Walter has
suggested for const could resolve a lot the current issues that have
been raised lately.
It sounds to me that if a class contains a shared member that the shared
member would remain non-const, even if the class reference becomes const
or invariant.
Other observations:
1. Having a read lock on a variable sounds equivalent to having a scope
invariant variable.
2. Having a write lock forces only one writer, a concept that is not
covered by Walter's const proposal.
Since I haven't really seen any motivating use cases for how the new
const would be used and I don't fully appreciate how Walter's proposed
design would truly satisfy the requirements, I'll stop my speculative
language design here.
More information about the Digitalmars-d
mailing list