Ok after 2 minutes more of thought I guess the response will most likely be "which integer will you be referencing in the result of the metafunction" -- so maybe it can be an error to use an unadorned pattern variable twice in a pattern?
On 05/01/2012 05:45 PM, Jon Rafkind wrote: > The default matching behavior of patterns in redex will try to match > duplicate pattern variables to the same term. Thus > > #lang racket > > (require redex) > > (define-language Z [x integer]) > (define-metafunction Z > simple : integer integer -> integer > [(simple integer integer) 1]) > > (term (simple 1 2)) >> simple: no clauses matched for (simple 1 2) > Won't match because the pattern (simple integer integer) will match the first > `integer' to 1, and wants to the second reference to `integer' to the same > thing. To get it to work I need to adorn the pattern variables with _1 and _2 > to make them unique. > > (define-language Z [x integer]) > (define-metafunction Z > simple : integer integer -> integer > [(simple integer_1 integer_2) 1]) > > I would prefer (if possible) if the default behavior treated unadorned > pattern variables as fresh so that I don't have to remember to attach a > suffix. I guess this might break existing programs but I can't imagine anyone > really relies on this behavior (unadorned variables matching duplicate terms). > ____________________ > Racket Users list: > http://lists.racket-lang.org/users ____________________ Racket Users list: http://lists.racket-lang.org/users
