Re: Order of cluster searching with ace file (recent snapshot)

Le Mercredi 13 D=E9cembre 2006 17:39, Cyril ADRIAN a =E9crit=A0:
> Hi Dominique,
>
> On 12/13/06, Colnet Dominique <[email protected]> wrote:
> > You just have to use a long distance (i.e. in a deep directory).
>
> I think I have found an example where distance is not enough: what if
> you have two libraries, say "banking" and "address book", that have a
> class PERSON (one pointing to their bank accounts; the other having
> the person's address), and you want to use both? One solution is to
> bridge both but I don't find it ideal since you need quite a complex
> setting:
>
> root/root.e -- uses BANK_PERSON and ADDRESS_BOOK_PERSON
>  + bridge_bank/bank_person.e -- facade of bank's PERSON
>
>  |  + bank/person.e
>
>  + bridge_address_book/address_book_person.e -- facade of address book's
> PERSON + address_book/person.e
>
> I don't find it ideal because the library's code needs to be nested
> *inside* the application tree :-(
>
> Any ideas?

maybe.

If only the loadpath.se were used, it could be simple to use it to=20
hierarchically search the classes.=20

=46or your example, in the subdirectory root/bridge_bank, the file loadpath=
=2Ese=20
would contain:

	.
	./bank

And in the subdirectory root/bridge_address_book, the file loadpath.se woul=
d=20
contain:

	.
	./address_book

Then for the classes in the subdirectory root/bridge_address_book, the used=
=20
classes would be searched first from root/bridge_address_book, then from=20
root/bridge_address_book/adress_book and then from previously stacked=20
loadpath including the defaults one.

The nested hierarchical scope of search would be depending of where is the=
=20
currently compiled class.

jos=E9