Subject: publish
From: Vassily Litvinov (vass@cs.washington.edu)
Date: Wed Mar 28 2001 - 17:30:58 PST
I have published parser and typechecker changes in Cecil/src/compiler
and Cecil/src/stdlib. My typechecker successfully typechecks vortex
and whirlwind. I plan to install new solaris and linux executables tomorrow.
SYNTAX and SYNTACTIC SUGARS
- clarified and corrected implementation of "isa"/"subtypes" declarations
w.r.t. backquote sugar (see end for details)
- type synonym declarations are introduced with the following syntax:
type synonym A standsfor T;
This declaration introduces a type name "A" which, when used, say in
let var x:A := ...;
will desugar into T, which must be a legal top-level type. I.e. all type
names inside T are lexically scoped under the top-level type environment
(as opposed, e.g., to nested inside some method; therefore, e.g., no type
variables are allowed as there are none at top level).
-- couldn't use "is" as keyword as one of the *existing* fields of
the synonym declaration object ironically is called "is"
NOTE: synonyms seem to work but haven't been tested extensively.
Report all bugs to me...
- object/type declarations cannot refer to synonyms or use backquote sugar,
to avoid circular dependencies. This does not apply to "extend"
declarations and to "isa"/"subtypes" clauses.
- synonym declarations cannot refer to synonyms, for same reason
-- this could be relaxed with more coding, for which I don't have time
- now I at least understand what the backquote sugar does
and btw now you can declare covariance of, say, i_vector,
without using an "extend" declaration
-- precise spec is due; meanwhile ask me if things behave weird
- new syntactic sugar: constraint "where sometype" stands for
"where sometype <= void" (recall "void" is the top type)
- likewise, "type synonym A;" stands for "type synonym A standsfor void;"
-- the reason for introducing these will be revealed later
MISCELLANEOUS
- in stdlib added cast_from[]() to allow programmer verify *from* what type
the value is being cast
- fixed a type error in stdlib
- got all typechecker files to typecheck
- rewrote and deleted a lot of code in typechecker; added a few comments
TO DO
- agree on syntax for type synonyms
-- I am very tempted to use "=" instead of "standsfor"
- allow "[object] synonym" along with "type synonym"
- implement desugaring of "@:" in specializers of method formals
-- currently "method foo(x@:my_synonym):void" is not supported
- update Cecil language spec
-Vass
=============================================================================
DETAILS
The following sugar is now properly implemented and is applied before
desugaring the backquote sugar:
declaration
forall F where W1: object O1 where W2 isa O2, O3 subtypes T4;
desugars into:
forall F where W1,W2: object O1;
forall F where W1,W2: extend object O1 isa O2;
forall F where W1,W2: extend object O1 isa O3;
forall F where W1,W2: extend object O1 subtypes T4;
declaration
forall F where W1: extend object O1 where W2 isa O2, O3;
desugars into:
forall F where W1,W2: extend object O1 isa O2;
forall F where W1,W2: extend object O1 isa O3;
where
F is a list of type variables, as before
W1, W2 are lists of constraints, as before
"where W1", "forall F where W1", and/or "where W2" may be omitted as before
The same scheme applies to similar declarations with "object" replaced by
"concrete object", "representation", "type", etc., and "isa" replaced
by "inherits" or "subtypes".
=============================================================================
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://majordomo.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2b25 : Wed Mar 28 2001 - 17:31:04 PST