Re: [abc] Defining used terms

On Wed, 29 Sep 2004, Aske Simon Christensen wrote:

>
> Hi!
>
> Throughout the paper, the terms 'name pattern' and 'classname pattern' are used
> a lot. The AspectJ manual only defines the terms 'type pattern' and 'type name
> pattern', so if we use any other terms, we should define them first. In any
> case, we should avoid using 'classname pattern' since it is quite misleading.
> It is not a pattern over names, since it contains the + (subtype) operator.

ok. We decided on "named class pattern expressions" where we talk about
the extension of [6].

> I looked for occurrences of 'shadow' as well, to check that we consistently
> speak of join point shadows (or this is clear from the context), which seems to
> be the case, except for once in joinpointsbasic (which I fixed) and in the
> intro, where it says:
>
> For instance, Kiczales has
> proposed a new type of pointcut, called {\em predicted cflow} \cite{kiczales03}.
> ${\bf pcflow}(p)$ matches at a shadow if there may exist a path to a
> another shadow where $p$ matches.
>
> Shouldn't it say 'join point' instead of 'shadow' in both places here?

I thought quite carefully about this. The existence of a path is relative
to a *static* notion of flow graph. Consequently, the matching in this
case is also a purely static notion, on shadows rather than actual
joinpoints (which are a runtime notion). Ganesh disagrees because he can
imagine an implementation where the prediction is done at runtime, and
he's going to change it.
Received on Wed Sep 29 14:51:48 2004