<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta content="text/html;charset=UTF-8" http-equiv="Content-Type"> <title></title> </head> <body bgcolor="#ffffff" text="#000000"> Matthew Flatt schreef: <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap="">Ok, but I'm still missing something, because I don't see how the table accomplishes anything. In particular, ekeeping the key maintains the mapping to the weak box, but it doesn't keep the weak box from getting emptied. </pre> </blockquote> Ok, another example session (see below). "A weak box is similar to a normal box (see section <a href="http://download.plt-scheme.org/doc/mzscheme/mzscheme-Z-H-3.html#node_sec_3.10">3.10</a>), but when the automatic memory manager can prove that the content value of a weak box is only reachable via weak boxes, the content of the weak box is replaced with <code class="scheme">#f</code>." As I see it, in the example below, a, b and c all point to the memory "contained" in the weak box. As long as a, b and c exist, the contents of the weak box will never be removed. And that is exactly what I want. You see, what happens is as follows: 1. psearch actually does a RPC via TCP/IP on localhost, port 4001. telnetting on localhost port 4001 gives the following session: <tt>(OODB "LPW0251.strader.xs4all.nl" 0.1 "oodb-backend-sqli" 0.1 1) (search (all (illness-class-category)) ())(ok ((59872449310226502864700443097271 illness-class-category))) </tt>2. The result: <tt>((59872449310226502864700443097271 illness-class-category)) </tt>indicates that OID 59...271 is of class illness-class-category. 3. Next, an object of class illness-class-category is instantiated with 'from-oid OID. This instantiates a class illness-class-category with default parameters. 4. Now I store (list OID) in the instantiated object and I put the object in a weak box. The (list OID) of object I use as key in the hash table. 5. When I retrieve another <tt>((59872449310226502864700443097271 illness-class-category)) </tt>from the server, I first look for (list OID) in the hash table. HEY, there's an object already there, I'll use this one instead of allocating a new one. Now, what I've effectivly reached, is that within the same running instance of an application, two objects of the same OID, which are eq? in the object oriented database (OODB) I've created, are also eq? in the running instance. And that's just what I wanted. Thanks to Eli's input! <tt>> (require "proos.scm") > (require "oodb-ci.scm") > (oodb-ci-initialize "localhost" "hans" "" 4001) #t </tt><tt>> (require "zk-classes.scm") > (define a (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #f #f PROOS-STORE: 59872449310226502864700443097271 #roos > (define b (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #<weak-box> #roos > b #roos > (eq? a b) #t > (-> a name) "ok" > (-> b name) "ok" > (-> a name "new name") ok-put! > (-> b name) "new name" > (collect-garbage) > (define b (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #<weak-box> #roos > (collect-garbage) > (define b (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #<weak-box> #roos > (collect-garbage) > (define c (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #<weak-box> #roos > c #roos > (set! a 0)(set! b 0)(set! c 4) > > > (define c (car (psearch (all illness-class-category)))) PROOS-STORE: 59872449310226502864700443097271 #<weak-box> #roos > c #roos > (-> c name) "new name" (-> c illness-classes) PROOS-STORE: 163698054800566253127307108927946 #f #f PROOS-STORE: 163698054800566253127307108927946 #roos PROOS-STORE: 163698054800566253127307108927945 #f #f PROOS-STORE: 163698054800566253127307108927945 #roos PROOS-STORE: 163698054800566253127307108927944 #f #f PROOS-STORE: 163698054800566253127307108927944 #roos PROOS-STORE: 134167358730453000620204068929124 #f #f PROOS-STORE: 134167358730453000620204068929124 #roos PROOS-STORE: 132069009480453002472005822925980 #f #f PROOS-STORE: 132069009480453002472005822925980 #roos PROOS-STORE: 132069009480453002472005822925979 #f #f PROOS-STORE: 132069009480453002472005822925979 #roos PROOS-STORE: 132069009480453002472005822925978 #f #f PROOS-STORE: 132069009480453002472005822925978 #roos PROOS-STORE: 101069891380339751816804499004492 #f #f PROOS-STORE: 101069891380339751816804499004492 #roos PROOS-STORE: 68254699550226382758901006667984 #f #f PROOS-STORE: 68254699550226382758901006667984 #roos PROOS-STORE: 63721520500226501175203246565420 #f #f PROOS-STORE: 63721520500226501175203246565420 #roos (#roos #roos #roos #roos #roos #roos #roos #roos #roos #roos) > (define b (car (-> c illness-classes))) PROOS-STORE: 163698054800566253127307108927946 #<weak-box> #roos PROOS-STORE: 163698054800566253127307108927945 #<weak-box> #roos PROOS-STORE: 163698054800566253127307108927944 #<weak-box> #roos PROOS-STORE: 134167358730453000620204068929124 #<weak-box> #roos PROOS-STORE: 132069009480453002472005822925980 #<weak-box> #roos PROOS-STORE: 132069009480453002472005822925979 #<weak-box> #roos PROOS-STORE: 132069009480453002472005822925978 #<weak-box> #roos PROOS-STORE: 101069891380339751816804499004492 #<weak-box> #roos PROOS-STORE: 68254699550226382758901006667984 #<weak-box> #roos PROOS-STORE: 63721520500226501175203246565420 #<weak-box> #roos > (proos-store->list) (((163698054800566253127307108927944) #<weak-box>) ((101069891380339751816804499004492) #<weak-box>) ((134167358730453000620204068929124) #<weak-box>) ((68254699550226382758901006667984) #<weak-box>) ((132069009480453002472005822925979) #<weak-box>) ((132069009480453002472005822925980) #<weak-box>) ((59872449310226502864700443097271) #<weak-box>) ((132069009480453002472005822925978) #<weak-box>) ((63721520500226501175203246565420) #<weak-box>) ((163698054800566253127307108927946) #<weak-box>) ((163698054800566253127307108927945) #<weak-box>)) > (collect-garbage) > (proos-store->list) (((59872449310226502864700443097271) #<weak-box>) ((163698054800566253127307108927946) #<weak-box>)) > (set! c 0) > (proos-store->list) (((59872449310226502864700443097271) #<weak-box>) ((163698054800566253127307108927946) #<weak-box>)) > (collect-garbage) > (proos-store->list) (((163698054800566253127307108927946) #<weak-box>)) > b #roos (-> b poid) 163698054800566253127307108927946 > (set! b 2) > (collect-garbage) > (proos-store->list) () > </tt> ********************************************************** as for the class definitions ********************************************************** <tt>(def-pclass (this (illness-class-category . args)) (default) (supers) (private) (persistent name z-order ) (public (define (illness-classes) (psearch (= (illness-class category-oid) (poid this)))) (define (add-illness-class obj) (-> obj category-oid (poid this))) ) (constructor) ) (def-pclass (this (illness-class . args)) (default) (supers) (private) (persistent _category category-oid name color-fg color-bg font ) (public (define (category . cat) (if (not (null? cat)) (begin (-> this _category (car cat)) (-> (car cat) add-illness-class this))) (-> this _category)) ) (constructor) ) </tt> <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap=""> Are `a' and `b' keys here? </pre> </blockquote> No, the're the (Persistent) ROOS objects. <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap=""> Ok, but is what's in the weak box --- an object or #f? </pre> </blockquote> As can be seen from the previous section. Still the object. <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap=""> Looking back at your problem description, maybe what I'm missing is that it's ok for for the weak box to be emptied, because this table implements a cache? </pre> </blockquote> Well it sort of creates a cache. As long as there are "living" objects of a certain OID, the cache won't be collected. When all objects of a certain OID cease to exist, it is collected. <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap=""> </pre> <blockquote type="cite"> <pre wrap="">However when I don't use the (list) around the oid, It won't work. Isn't there a way without the 'equal property on the hash table? </pre> </blockquote> </blockquote> Ok, but then I'll have to convert a number to a symbol each time. I wonder what is more expensive? Note that the numbers I use are fairly large. Doesn't that open possibilities? <blockquote cite="mid200511221430.jAMEU3dq029877@pegasus" type="cite"> <pre wrap="">Using symbols instead of numbers might be the right trade-off. Symbols are allocated and collected, like lists, but they are also interned, like small integers. Matthew </pre> </blockquote> Thanks, Hans </body> </html>