[abc] an observation about annotation patterns

When I was playing around with grammars for annotation patterns
yesterday, the conclusions I came to is that the difficult part is not
the grammar for the patterns themselves, but rather deciding all the
places that they can go. In particular, it is *not* true that we just
replace every type pattern by a pair @annotation type.

Example: Currently, we have !Bar. But in !@Foo Bar, the ! means
something completely different.

Example: Currently, we have Class.method(). But in @Foo Class.method(),
apparently the @Foo applies to the method, not the Class.

So it seems that there will be many places in the grammar where an
annotation will be allowed. Our next task will be deciding on what all
those places are.

Of course, probably in all or most cases, a type pattern can be replaced
by (@annot type). But in some cases, the parentheses are likely to not
be required, while in other cases, they will be.

Ondrej
Received on Tue Dec 7 11:57:31 2004