On a sunny Monday afternoon, Michael Brouwer sent this opinion to the TOM mailing list: In response I changed the syntax of getting a
selector into what he suggested. If you never need to write
anything in C, you can safely forget about the old syntax.
However, for glueing TOM and C, you'll
probably need to know how selectors are handled internally.
Examples
Suppose you have this object a, which implements the
following (silly) method:
|
int (result)
tell Foo object
toMultiply int p
by int q;
|
An invocation of this method is straightforward:
|
[a tell b toMultiply 6 by 9];
|
The same invocation, but going through perform could
look like this:
|
[a perform selector (int tell Foo toMultiply int by int)
with (b, 6, 9)];
|
Thus, the selector lists the return type, followed by the method's
name parts and argument types. Put differently: the stuff between
the parenthesis is like a method heading, but without any argument
names. To clarify: compare the original method declaration above
with the selector written like this:
|
selector (int
tell Foo
toMultiply int
by int)
|
Do not be fooled by the Foo type in the selector; it
is the same selector as:
|
selector (int tell All toMultiply int by int)
|
or
|
selector (int tell MutableArray toMultiply int by int)
|
because for selectors, there is no difference between objects,
just like the TOM semantics for method overloading do not discern
between different object classes (all objects are objects). On
the other hand, denoting the object type in the selector is great
for documentational purposes.
Small print
In TOM, the syntax type
(expression) is used for type casts.
Example: long a = 0x1234567890; int b = int (a);
Because of the new selector syntax, the selector type
can no longer be used in type casts. Not that it really matters,
since the only thing that can be cast to a selector is a selector,
which is not a very useful operation. It is, however, an
irregularity, albeit one of which the advantages outweigh the
disadvantages.
Up: Highlights
|