New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Unexpected result of set-contains? #400

Closed
zephyrfalcon opened this Issue Dec 1, 2018 · 2 comments

Comments

Projects
None yet
2 participants
@zephyrfalcon
Copy link

zephyrfalcon commented Dec 1, 2018

Hi,

I am seeing unexpected results when creating a set, adding elements with set-adjoin and then checking if they are in the set with set-contains?. This is my code:

(use scheme.set)

(define my-set
  (let loop ((a-set (set eq-comparator))
             (numbers (iota 20)))
    (if (null? numbers)
        a-set
        (loop (set-adjoin a-set (car numbers)) (cdr numbers)))))

(print (set-size my-set))   ;; length is correct
(print (set->list my-set))  ;; everything shows up

;; some of these show up as #f!
(set-for-each
 (lambda (x)
   (print "Set contains " x "? " (set-contains? my-set x)))
 my-set)

I am getting these results:

20
(14 15 16 17 12 13 18 19 1 0 11 10 9 8 7 6 5 4 3 2)
Set contains 2? #f
Set contains 3? #f
Set contains 4? #f
Set contains 5? #f
Set contains 6? #t
Set contains 7? #t
Set contains 8? #f
Set contains 9? #f
Set contains 10? #f
Set contains 11? #f
Set contains 0? #t
Set contains 1? #t
Set contains 19? #t
Set contains 18? #t
Set contains 13? #t
Set contains 12? #t
Set contains 17? #t
Set contains 16? #t
Set contains 15? #t
Set contains 14? #t

As far as I can tell, set-size works properly, so does set->list which shows all the elements I added. But when I loop over the set with set-for-each, and test if each element is in the set (the same set we're looping over), then set-contains? returns #f for a number of elements.

The actual results may be different on other computers, I don't know. When I use 10 elements, they all show up. With 20, I am starting to see the above.

If I create the set with (list->set eq-comparator (iota 20)), it does work as expected. Am I using set-adjoin incorrectly, or is this a bug?

I am using Gauche 0.9.6 (Homebrew, latest version) on macOS Sierra (10.12.6). To be precise, Gauche scheme shell, version 0.9.6 [utf-8,pthreads], x86_64-apple-darwin16.7.0.

@shirok

This comment has been minimized.

Copy link
Owner

shirok commented Dec 1, 2018

Wow. Confirmed on Linux and development HEAD. That's perplexing, but something seems fishy in the underlying hashtable. Looking at it.

shirok added a commit that referenced this issue Dec 1, 2018

Fix: hash-table-copy that didn't preserve entries' hashval
The bug was lurking unnoticed since entries' hashval is only used
when hashtable is rehashed.

The cause of #400
@shirok

This comment has been minimized.

Copy link
Owner

shirok commented Dec 1, 2018

Turned out this is an old bug in hash-table-copy that hadn't been noticed (amazingly!). It only surfaces when you copy a hashtable then add a bunch of entries to the copy so that it triggers rehashing. I guess it haven't been noticed since we've been using hashtables as mutable object for most of the time. The set interface is mostly functional, so internally it copies hashtables, hence it finally came up. Thanks for reporting!

@shirok shirok closed this Dec 13, 2018

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment