Check-eq? not behaving properly in Typed Racket

Questions & Answers
1

Consider:

#lang typed/racket/base

(require typed/rackunit)

(: hash->immutable-hash (All (k v) (-> (HashTable k v) (Immutable-HashTable k v))))
(define (hash->immutable-hash htab)
  (if (immutable? htab)
      htab
     '(more code here)))

(define h '#hasheq((a . 1) (b . 2) (c . 3)))
(eq? (hash->immutable-hash h) h) ; => #t
(check-eq? (hash->immutable-hash h) h) ; Fails.

Basically, if passed an immutable table, it's returned, otherwise it creates a new immutable table with the same contents as the original.

When testing the "return the argument if already immutable" case, I can't figure out why check-eq? fails when the equivalent eq? succeeds.
The same test case passed in an standard untyped Racket implementation; it's just in TR that there's a problem.

2

Hi, @shawnw.

I can't really comment on the why, but it seems like it is a known issue, at least, that Typed Racket and RackUnit don't play well together in this regard.

Sorry for the nothingburger.

3

An eight year old thread and the issue is still there, yay. (check-true (eq? ... ...)) seems to work though.

1 Like
4

The explanation is simple. The types compile to contracts before values escape to untyped code, possibly a library. Some contracts wrap values in a way that obscures pointer equality.

I bumped into this issue myself over the summer when I created a TR benchmark from a little project I wrote. I went back and questioned why I used eq? and not equal? -- @robby 's old question from our 2004 paper. No I didn't need to use eq.