Things work now as I expect for weak boxes, but I am still confused about weak hash tables. I do the following: #lang racket (define h (make-weak-hash)) (define keys (for/list ((key (in-list '(a b c d e f)))) (make-weak-box key) ; <--- 1 ; key ; <--- 2 )) (define (enter key value) (will-register will-executor value will-proc) (set! enter-counter (add1 enter-counter)) (hash-set! h key value)) (define-syntax-rule (while condition expr ...) (let loop () (when condition expr ... (loop)))) (define will-executor (make-will-executor)) (define (will-proc v) (set! release-counter (add1 release-counter)) #t) (define enter-counter 0) (define release-counter 0) (define range (in-range 6)) (for ((key (in-list keys)) (value range)) (enter key (list value))) (list enter-counter release-counter) ; -> (6 0) ok h ; -> hash with 6 entries, ok (set! keys #f) ; discard all keys (collect-garbage) (collect-garbage)(collect-garbage) (collect-garbage) (while (will-try-execute will-executor)) (list enter-counter release-counter) ; -> (6 6) ok h ; -> empty hash, ok So far ok. However, when commenting out line <--- 1 and uncommenting line <--- 2, I would expect the same results, but I find: (6 0) ; ok hash with 6 entries ; ok (6 0) ; a surptrise for me hash with 6 entries ; a surprise for me This is the cause of my confusion. Surely I am misinterpreting the docs. Help much appreciated. Thanks again, Jos
_____ From: Carl Eastlund [mailto:[email protected]] Sent: martes, 28 de enero de 2014 1:28 To: Jos Koot Cc: Racket Users Subject: Re: [racket] FW: weakly held symbols? This isn't a typo. The docs are talking about the properties of the symbol tables for interned and unreadable symbols. There is no symbol table for uninterned symbols, so it wouldn't mean anything to talk about it. Here's what this means for an interned symbol: you can create one, and while a reference to it exists, it stays in the symbol table. Any time you intern the same string, you'll get back that symbol. However, if all references to that symbol vanish, the symbol table is free to release that symbol. If anyone interns the same string again later, a new symbol will be created. For nearly all purposes, this is completely transparent, because you couldn't possibly have a copy of the old symbol to compare. There's almost no way to tell that the symbol you got before and the one you got after were different. The only way you'd ever know would be either by tracking memory use statistics, or by comparing the results of something like eq-hash-code, neither of which should be surprising to find out behaves impurely. Carl Eastlund On Mon, Jan 27, 2014 at 5:53 PM, Jos Koot <[email protected]> wrote: It appears that my previous email was not sent completely. Sorry for that. Here is the complete question. Hi to all, Section "3.6 symbols" of the Racket reference manual states: Interned and unreadable symbols are only weakly held ... but a symbol maay disappear when ... used as the key in a weak hash table ...". I would understand: UNINTERNED and unreadable symbols are only weakly held ... but a symbol may disappear when ... used as the key in a weak hash table ...". May be just a typo, but as I am not sure I understand the docs well, I post my question here instead of posting a bug report. Greetings, Jos ____________________ Racket Users list: http://lists.racket-lang.org/users
____________________ Racket Users list: http://lists.racket-lang.org/users
