abbott@aerospace.aero.org (Russell J. Abbott) (09/08/89)
If anyone is interested I've discovered a cute way to implement a "list"
operation in smalltalk. If this is old hat, excuse my mentioning it.
In Lisp, one can write, for example,
> (list (plus 1 2) 4)
and get
(3 4)
That is, "list" takes an arbitrary number of arguments and returns a
list containing the evaluations of those arguments. In smalltalk there
appears to be no equiivalent. One could write something like:
#( (1 + 2) 4)
but one would get an array containing two elements, the first of which
is an array containing the three elements: 1, +, and 2. That is, there
is no evaluation.
The way around this is to define:
(1) an instance creation method (which I call "<") that inserts
its argument into the newly created object and
(2) an "add:" method (which I call ",")
for Collection. Then one can write:
Set < (1 + 2), 4
This will yield: Set(3 4). The "<" operator should be read similarly to
its use in the Unix shell as "redirect," i.e., create a set and put
these elements into it. Of course, this structure may be nested:
Set < 1, (Set < (2+3), 4), 5
which yields: Set(1, Set(5, 4), 5)
While doing this, I found myself wishing that smalltalk had a parser
similar to prolog's in which one could define and redefine operators and
their precedences. The smalltalk syntax is so simple, such a parser
should not be too much to ask.
--
-- Russ abbott@itro3.aero.orgjohnson@p.cs.uiuc.edu (09/12/89)
Russ Abott's list operation is very cute. One potential problem is that "," is already defined in SequenciableCollection to concatenate two collections. It is real handy with strings.
abbott@aerospace.aero.org (Russell J. Abbott) (09/13/89)
In article <80500077@p.cs.uiuc.edu> johnson@p.cs.uiuc.edu writes: > >Russ Abott's list operation is very cute. One potential problem is >that "," is already defined in SequenciableCollection to concatenate >two collections. It is real handy with strings. For just that reason I now use "/", e.g., Set / 1 / 2 / 3 / 4 ... . The "/" message is defined at both the class and the instance levels. Of course one can pick one's own operator symbol. -- -- Russ abbott@itro3.aero.org