prj@po.cwru.edu (Paul Jarc) writes:

> How about this:
> (let* ((recipients '(("foo@bar" . "Name <foo@bar>")
>                      ("baz@quux" . "None <baz@quux>")))
>        (plain-addrs (mapcar 'car recipients))
>        (gnus-hierarchical-lists '(("baz@quux" "foo@bar")))
>        sublists recip)
>   (while plain-addrs
>     (setq sublists (assoc (car plain-addrs) gnus-hierarchical-lists)
>           plain-addrs (cdr plain-addrs))
>     (when sublists
>       (setq sublists (cdr sublists))
>       (while sublists
>         (setq recip (assoc (car sublists) recipients)
>               sublists (cdr sublists))
>         (if recip
>             (setq recipients (delq recip recipients)))))))
>
> Note that this is a different structure for gnus-hierarchical-lists.
> An entry (A B C) means that A contains B and C.  (No further
> relationship between B and C is implied.)  So if A appears in the
> recipient list, then B and C will be removed from it, if present.  If
> B and C have sublists of their own, they should get their own
> top-level entries in gnus-hierarchical-lists.  In this case,
> gnus-hierarchical-lists says that messages to baz@quux also go to
> foo@bar, so foo@bar is removed from the recipient list.

That looks fine to me.

Kai, sorry for not following up to your very helpful message.  I was
unable to send e-mail to the ding list for a while due to server
problems.  I think Paul's suggestion is even better to make the user's
life simple, using a list instead of a cons cell.

I tried to put a patch into message.el, naming the variables
accordingly and simplifying things a bit.  I'm not sure if the
variable should go into the message-headers or the message-sending
group; also, I'm not sure if I defined the variable correctly with
defcustom (I tried the '(repeat sexp) type but that didn't work).  I
think the functional definition, at least, is correct (because Paul
wrote it :)

I did a diff -u, renaming the original message.el to
message-original.el because I don't have access to the CVS repository
(firewall issues, don't ask) to do a diff against it.

This is my first time submitting a patch, please let me know if I did
anything wrong.  It's also my first time writing more than 4 lines of
Lisp for a single purpose.

To test, define message-hierarchical-lists to be ("foo" "bar") and
then have "foo" and "bar" in your recipient list.  "bar" should be
removed.

I couldn't test - message-get-reply-headers was not invoked, according
to debug-on-entry.  I'm kind of puzzled by this.  I must be doing
something wrong.

Also, the message-hierarchical-lists should probably be generalized to
regular expressions for both the primary and secondary lists
(message-hierarchical-lists-regexps?).  I don't need this personally,
but perhaps someone else will find it useful.  Maybe when the string
exact matching is working.

Thanks
Ted