Thanks, didn't know I could do that.
--- In
[hidden email], "pa_bates80" <paul.bates@...> wrote:
>
> Not at all. In fact you code could have had a serious bug because the "hidden" creation routine gets forgotten and some attributes may not be initialized correctly.
>
> The solution is to declare the creation routine exported to the containing class only.
>
> ...
>
> create
> make
>
> create {SOME_CLASS}
> some_creation_feature
>
> ...
>
> This will also satisfy compiler when using Void-Safe when checking all attached attributes have been correctly set.
>
> Paul.
>
> --- In
[hidden email], "Jimmy Johnson" <boxer41a@> wrote:
> >
> > I found an answer to my previous question at docs.eiffel.com "Converting existing software to void-safety". The example shows a problem just like mine:
> > 1. an ancestor has an entity `val' declared "attached like Current".
> > 2. there is a call "create val.some_creation_feature".
> >
> > This works fine for this class but doesn't this *force* descendents to add `some_creation_feature' to its list of creation features whether it wants to or not? This *has* to violate some principle, perhaps "information hiding" since I can no longer hide `some_creation_feature'.
> >
> > Also, why did that pattern work in the past but now is being rejected?
> >
> > Jimmy J. Johnson
> >
>
