-<xlog server="chat.freenode.net" channel="#existdb" date="2008-05-17">

<message nick="wolf77" time="11:09:07.991+01:00">Hi Perig</message>

<message nick="perig" time="11:09:40.595+01:00">Well, I think we'll have to plug the index onto max and min :-)</message>

<message nick="wolf77" time="11:10:44.132+01:00">I have been thinking about it for a while, but I see many problems.</message>

<message nick="perig" time="11:11:13.251+01:00">We could use the common way but it's probably not the good way to go. Also thinking...</message>

<message nick="perig" time="11:11:34.74+01:00">max((1 to 100)) is a good example </message>

<message nick="perig" time="11:12:19.261+01:00">somehow, the range expression should indicate it can collaborate with max() or... vice-versa</message>

<message nick="wolf77" time="11:12:39.318+01:00">the expression within min/max could be quite complex, which max it hard to determine what to index</message>

<message nick="wolf77" time="11:13:05.824+01:00">Also, it would be good if the same index facility could be used with "order by" as well</message>

<message nick="perig" time="11:13:20.023+01:00">is it *basically* so different from GeneralComparison ?</message>

<message nick="perig" time="11:13:39.76+01:00">OK, for the "order by" concern,of course</message>

<message nick="perig" time="11:14:43.276+01:00">other topic : a NPE in the ancestor axis ? That's bad...</message>

<message nick="wolf77" time="11:15:00.79+01:00">mmmh, the question is: where is the real bottleneck? is it in atomization? in comparisons?</message>

<message nick="wolf77" time="11:15:27.446+01:00">I would need to create a test case and run it through the profiler</message>

<message nick="wolf77" time="11:16:02.084+01:00">without that it's hard to say what approach would be most efficient</message>

<message nick="wolf77" time="11:16:44.081+01:00">I did not see the NPE in any of my applications so far</message>

<message nick="perig" time="11:16:49.047+01:00">I think the atomization (at least in this case)</message>

<message nick="wolf77" time="11:16:52.873+01:00">we really need a test case here</message>

<message nick="perig" time="11:17:15.744+01:00">Yeah : notice that it occurs during an optimization</message>

<message nick="wolf77" time="11:17:39.986+01:00">Ok, but if the bottleneck is in atomization, you would not safe much by creating an index</message>

<message nick="wolf77" time="11:18:14.363+01:00">after all, we have a simple attribute lookup here. no complex computation of string values. just one dom.dbx access.</message>

<message nick="wolf77" time="11:20:52.047+01:00">well, a simple approach could be to pre-compute lists of nodeId/atomicValue pairs in a special NodeSet implementation, then store them into a plain binary file on disk</message>

<message nick="perig" time="11:21:16.33+01:00">sounds good</message>

<message nick="wolf77" time="11:21:31.621+01:00">the disadvantage would be that you need to re-compute the node set after each document update</message>

<message nick="perig" time="11:22:27.177+01:00">I wonder if we shouldn't introduce some"dirty" flags after updates</message>

<message nick="DiZzZz" time="11:22:44.14+01:00">morning guys</message>

<message nick="wolf77" time="11:23:13.209+01:00">I already use a dirty flag somewhere... in elements.dbx I think</message>

<message nick="perig" time="11:23:57.459+01:00">BTW, I have a pending patch that gets rid ofLocation.setPreloadNodeSets(boolean doPreload) in favour of setPreloadedData(DocumentSet docs, NodeSet nodes) only.OK to commit ?</message>

<message nick="perig" time="11:27:04.223+01:00">These preloaded node sets are still unclear to me.</message>

<message nick="wolf77" time="11:28:00.599+01:00">There are two reasons for using those:</message>

<message nick="wolf77" time="11:28:09.231+01:00">1) in a nested for loop</message>

<message nick="wolf77" time="11:28:39.763+01:00">for $a in //foo let $b := $a/baz return $b</message>

<message nick="wolf77" time="11:29:06.178+01:00">it often is cheaper to preload all $a/baz during the first iteration</message>

<message nick="perig" time="11:29:52.822+01:00">Hi Dizzzz BTW</message>

<message nick="adrian111" time="11:29:58.91+01:00">hello...i have a question about exist...i wanna parse a file stored in existdb get a value of a field and change this value...could someone please give me some hints? :)</message>

<message nick="wolf77" time="11:30:07.181+01:00">2) the query optimizer uses setPreloadedData to pass in pre-computed results</message>

<message nick="perig" time="11:30:52.593+01:00">yes, I understand that. my question is rather : when should wepreload and when shouldn't we ?</message>

<message nick="perig" time="11:31:13.511+01:00">adrian1111 : there are nofields in XMLfiles ;-)</message>

<message nick="perig" time="11:31:50.452+01:00">as long as you'll stick to this concept, you won't get the picture</message>

<message nick="adrian111" time="11:31:55.418+01:00">sorry...tags :)</message>

<message nick="perig" time="11:32:39.969+01:00">neither...</message>

<message nick="wolf77" time="11:33:23.939+01:00">as long as we don't have more statistics on node distributions, we should always preload if there's a dependency on an outer for loop</message>

<message nick="perig" time="11:33:25.25+01:00">nodes and only nodes</message>

<message nick="adrian111" time="11:33:53.028+01:00">sorry...i'm kind of knew :)</message>

<message nick="perig" time="11:33:59.727+01:00">wolf77 : thanks for the clarification.Even if one disagrees, itis clear :-)</message>

<message nick="perig" time="11:36:45.813+01:00">maybe a comment in the code would help. Well, we all should do that :-/</message>

<message nick="perig" time="11:42:10.534+01:00">I have another patch bor baseURI but I can't remember why I've written it:-)</message>

<message nick="perig" time="11:42:43.639+01:00">(and still another one for util:eval)</message>

<message nick="wolf77" time="11:45:42.782+01:00">ok, lunch time ...</message>

<message nick="wolf77" time="13:05:10.728+01:00">probably yes</message>

<message nick="wolf77" time="13:05:28.542+01:00">did you run the performance tests?</message>

<message nick="wolf77" time="13:05:39.437+01:00">my laptop is too slow for it</message>

<message nick="adrian111" time="16:58:15.624+01:00">hello</message>

<message nick="adrian111" time="16:59:14.223+01:00">wolf : i want to change an xml file atribute stored in exist with a script function...do i need some kind of permision to do that?</message>

<message nick="adrian111" time="21:03:26.033+01:00">hello...i'm trying to modify an attribute of a xml file stored in exist with javascript code...it won't work...is there anything else i have to do besides setting the write permissions?</message>

<message nick="perig" time="21:04:46.212+01:00">tell us more about "it won't work" ? :-)</message>

<message nick="adrian111" time="21:14:39.087+01:00">i parsed this xml file with DOM and i cannot seem to replace an aribute</message>

<message nick="adrian111" time="21:14:47.038+01:00">attribute</message>

<message nick="perig" time="21:15:16.558+01:00">Errr... in your DOM or in the database one ???</message>

<message nick="adrian111" time="21:15:53.208+01:00">the file is stored in an exist collection</message>

<message nick="adrian111" time="21:16:00.638+01:00">i loaded the file via URL</message>

<message nick="adrian111" time="21:16:31.77+01:00">i got the value of the attribute that i'm suposed to change</message>

<message nick="perig" time="21:17:10.042+01:00">what lets you think that changing your DOM will be reflected to the DB ?</message>

-<message nick="adrian111" time="21:17:15.77+01:00">

setAttribute("url","

<a href="http://localhost:8080/exist/rest/db/arheo/HelloWorld1.x3d">http://localhost:8080/exist/rest/db/arheo/HelloWorld1.x3d</a>

");

</message>

<message nick="adrian111" time="21:17:51.048+01:00">i changed the dom and saved xmlDoc</message>

<message nick="perig" time="21:18:28.018+01:00">and... it doesn't work ?</message>

<message nick="adrian111" time="21:18:56.647+01:00">i gave full permission to that file</message>

<message nick="adrian111" time="21:19:41.718+01:00">of exist...let me take a look</message>

<message nick="adrian111" time="21:20:09.686+01:00">maybe i'm not saving the file right...</message>

<message nick="perig" time="21:20:10.507+01:00">well, you should have had a look before asking ;-)</message>

<message nick="adrian111" time="21:20:24.606+01:00">xmlDoc.load(link);</message>

<message nick="adrian111" time="21:20:31.806+01:00">this is the loader</message>

<message nick="adrian111" time="21:20:43.562+01:00">xmlDoc.save(link);</message>

<message nick="perig" time="21:24:15.92+01:00">Wel, lI don't know how these methods are implemented. Also, I don't know what is your good reason for doing this in JavaScript.</message>

<message nick="adrian111" time="21:25:00.03+01:00">i want to create a dynamical scene graph</message>

<message nick="adrian111" time="21:25:08.712+01:00">i'm storign x3d files</message>

<message nick="adrian111" time="21:25:14.53+01:00">storing</message>

<message nick="adrian111" time="21:26:17.786+01:00">and i have an x3d file - the main scene which has inline url to the other x3d files stored in exist...and i want to be able to import new objects in the scene in real time :)</message>

<message nick="adrian111" time="21:29:39.77+01:00">i don't have any exist error in the log</message>

<message nick="perig" time="21:29:53.919+01:00">yes, sure. But I still can't see the imperious need for JavaScript in order to change an attribute value</message>

<message nick="adrian111" time="21:30:57.235+01:00">any suggestions are more than welcomed :)</message>

<message nick="perig" time="21:31:01.441+01:00">I think there is no error in this log because eXist doesn't know anything about what you are trying to achieve; </message>

<message nick="perig" time="21:31:14.629+01:00">At least, this is my first guess.</message>

<message nick="adrian111" time="21:31:34.186+01:00">for example u have to add a new node to an xml file...how do u do that in code</message>

<message nick="perig" time="21:31:46.202+01:00">Anyway, the creation of a new document or the update of an existing one should be logged somehow</message>

<message nick="perig" time="21:32:19.958+01:00">I'd use an XQuery statement, or an XUpdate</message>

<message nick="adrian111" time="21:33:22.803+01:00">yes but i haven't got any experience with these...</message>

<message nick="adrian111" time="21:33:39.886+01:00">and i have a deadline with his</message>

<message nick="adrian111" time="21:34:33.189+01:00">which one would u use</message>

<message nick="perig" time="21:34:51.654+01:00">XQuery</message>

<message nick="adrian111" time="21:35:17.158+01:00">ok thanks for you answers :)</message>

<message nick="perig" time="21:35:34.652+01:00">I hope you can proceed now:-)</message>

<message nick="adrian111" time="21:35:45.257+01:00">i'd better :))</message>

<message nick="adrian111" time="21:36:42.985+01:00">goodbye and thanks again</message>

<message nick="perig" time="21:36:46.93+01:00">Frankly, why did you consider JavaScript ?</message>

<message nick="adrian111" time="21:37:27.776+01:00">well i've worked before...and since i'm creating a web application i thought it would be easier to develop</message>

<message nick="adrian111" time="21:38:45.361+01:00">i parsed xml and x3d documents before but i never had to modify one that is stored in a database</message>

<message nick="perig" time="21:38:53.492+01:00">We seldom mention JS on our web site...</message>

<message nick="adrian111" time="21:39:45.593+01:00">i know...but i manage to connect to exist...to load x3d files in players...</message>

<message nick="perig" time="21:39:49.028+01:00">We have a comprehensive documentation on how to modify a stored document.</message>

<message nick="adrian111" time="21:40:09.776+01:00">could you point out that doc please</message>

<message nick="perig" time="21:41:06.713+01:00">errr... ask Google ? read the mailing-list ?</message>

<message nick="adrian111" time="21:48:44.885+01:00">have a good day / night </message>

</xlog>