From: Wolfgang M. <me...@if...> - 2003-11-25 16:36:10
|
Hi, I just checked in my recent changes, which add support for some more XQuery/XPath functions, for example, fn:matches, fn:tokenize, fn:exists... Due to the updated Cocoon, several new problems with xalan, xerces and Cocoon showed up and I tried to fix these. Also, all dependencies on the fastutil library have been removed. fastutil provided hashtables and trees for native Java types. Unfortunately, fastutil was distributed in one, very large jar and used the C pre-processor to generate type specific code. Thanks to the help of my collegue, Stephan, we have been able to replace fastutil by our own simple hashtable package. The database may use some hashtables many thousand times a second, so having control over memory consumption, object removal and the like is quite important. Finally, the database now automatically determines default buffer sizes for each of the internal caches, so specifying buffer sizes in conf.xml is no longer necessary. Instead, you just define the maximum (raw) amount of memory to be shared by all caches. A typical entry thus looks like this: <db-connection database="native" files="webapp/WEB-INF/data" pageSize="4096" cacheSize="64M" free_mem_min="10"> cacheSize specifies the maximum raw amount of memory used if all caches are full (which is quite unlikely for most applications). However, the calculation assumes that each page consumes pageSize bytes. In reality, you have to add some bytes for Java object instance fields and things like that. As before, it is still possible to specify larger buffer sizes for selected caches, for example, to enlarge the cache for the fulltext index. There's also a new snapshot release, available at http://prdownloads.sourceforge.net/exist/eXist-snapshot-031125.zip?download Best regards, Wolfgang -- Wolfgang Meier Darmstadt University of Technology Tel.: +49 (0) 06151 16 4458 ICQ: #296595621 http://exist-db.org http://www.sozionet.org |