From: Steven Sharp (sharp@cadence.com)
Date: Sat Feb 12 2000 - 18:55:53 PST
> The latest proposal doesn't indicate whether attribute attachment is done
> as "zero or more" or "zero or one". In the BNF which of the following
> should be used?
> [ attribute_instance ]
> or
> { attribute_instance }
>
> I would like to see the latter, with the behavior being that all of the
> attributes are attached with the later one in the text having priority,
> for the case of "(* attr=1 *) (* attr=2 *)". But, if there are objections
> to this, I won't put up much of a fight.
I don't particularly care. Depending on how they are put into the BNF, some
productions are likely to allow multiple attributes via right-recursion,
even if square brackets are used. Allowing multiple instances on purpose
would avoid having to prevent them from being allowed by accident.
The effect of attaching multiple attribute instances would presumably be
identical to attaching a single attribute instance containing all of the
attribute definitions in order. One advantage I can see to allowing multiple
attribute instances is the ability to write macros that expand to self-
contained attribute instances, and apply more than one of them to an item.
Or the similar situation of specifying an additional attribute before an
include file that contains a module that already has its own attributes.
This archive was generated by hypermail 2.1.4
: Mon Jul 08 2002 - 12:54:12 PDT
and
sponsored by Boyd Technology, Inc.