CODEWITHSUNDEEP
racket
simmone
simmone

Reputation: 379

Racket module contract confuse me

the module: test-define.rkt

#lang racket

(provide test)

(provide (contract-out [add-test! (-> void)]))

(define test 0)

(define (add-test!)
  (set! test (add1 test)))

the main program:act.rkt

#lang racket

(require "test-define.rkt")

(printf "~a~%" test)

(add-test!)

(printf "~a~%" test)

run the act.rkt, I get:

0
1

this is what I want.

But if I change the contract in test-define.rkt:

(provide test)

change to 

(provide (contract-out [test integer?]))

then I run the act.rkt again, I get:

0
0

Why? I can't change the test value.

If I provide a get func, it turns normal again.

(provide (contract-out [get-test (-> integer?)]))

(define (get-test)
  test)

If test's type change to hash map, it's always normal.

What I missed?

Upvotes: 3

Views: 245

Answers (1)

soegaard
soegaard

Reputation: 31147

I notice that in test-define.rkt you have this line

(set! test3 (add1 test))

Should test3 be test ?

This might explain why you saw that two zeros (test never changed).

EDIT 2

For convenience I put your two modules in the same file and changed the contract of test:

#lang racket
(module test-define racket 
  (provide test)
  ; (provide (contract-out [test integer?]))
  (provide get-test)
  (provide (contract-out [add-test! (-> void)]))
  (define test 0)
  (define (add-test!)
    (set! test (add1 test)))
  (define (get-test)
  test))

(module ack racket
  (require (submod ".." test-define))
  (printf "~a~%" test)
  (add-test!)
  (printf "~a~%" test))

(require (submod "." ack))

Now I see the 0 1 vs 0 0 output as you do.

Hmmm. Why?

Well. If we change the provide forms to use no contracts at all, the output is 0 1.

Adding a contract shouldn't change this behaviour (I think). Maybe a bug?

http://pre.racket-lang.org/docs/html/guide/contracts-gotchas.html?q=contract&q=ignore

Says:

The contract library assumes that variables exported via contract-out are not assigned to, but does not enforce it. Accordingly, if you try to set! those variables, you may be surprised. ...snip... Moral: This is a bug that we will address in a future release.

Upvotes: 4

Related Questions