Reddit: SafeD - The Safe Subset of D
Georg Wrede
georg at nospam.org
Tue Mar 25 18:01:17 PDT 2008
Walter Bright wrote:
> I suspect that having a granular level of specifying safe/unsafe is the
> wrong approach. Doing it at the module level is easy to understand, and
> has the side effect of encouraging better modularization of safe/unsafe
> code.
That would mean that modules in Phobos would be either Safe or Unsafe.
Or[/and] that some modules would have to have two versions, one Safe and
the other Unsafe.
More practical would be (especially if the compiler has access to
info/hints to the safety of individual functions) to have it per function.
Then the compiler could discriminate, depending on if the user had used
the -Safe switch or not.
Personally, I'd advocate having Safety on App Level. Either an app is
SafeD compliant, or not. I have a hard time seeing anything in between.
More information about the Digitalmars-d-announce
mailing list