SV: [jdom-interest] Best strategy for caching JDom Document instance and provide concurrent read access to it?

Per Norrman pernorrman at telia.com
Thu Jan 8 14:55:55 PST 2004


Hi,

did something similar about two years ago, and i can't remember
concurrent read access being a problem. However, in this case
the motivation for using a cache was the time saved not parsing
a document.

If you need to duplicate an instance, I would definitely go with 
deep cloning.

As for the memory issue, as a rule of thumb I use a factor of 10, i.e.
memory usage = 10 x document size, but this varies a lot in real life.

/pmn


> -----Ursprungligt meddelande-----
> Från: jdom-interest-admin at jdom.org 
> [mailto:jdom-interest-admin at jdom.org] För Guillaume Berche
> Skickat: den 8 januari 2004 16:00
> Till: jdom-interest at jdom.org
> Ämne: [jdom-interest] Best strategy for caching JDom Document 
> instance and provide concurrent read access to it?
> 
> 
> Hello,
> 
> I'm very pleased with JDom API because it's simple and 
> intuitive. Thanks Jason for this great library! I looked into 
> the FAQ and into this list archive but could not find a 
> definitive answer to my question. Please point me to it if I 
> missed it.
> 
> 
> I'm trying to use the use-case described into the FAQ:
> "Single thread reads an XML stream into JDOM and makes it 
> available to a run time system for read only access"
> 
> Actually, I am trying to have a cache of JDom trees, and from 
> which a same JDom document instance may be access in read 
> only mode by concurrent threads.
> 
> In this list Jason wrote the following in 
> http://www.servlets.com/archive/servlet/ReadMsg?msgId=157461&l
> istName=jdom-i
> nterest.
> 
> "> JDOM is generally not thread safe, as I understand it.
> 
> True.  We follow the same model as ArrayList, which is not by 
> default thread safe."
> 
> 
> However ArrayList is actually safe for concurrent reads 
> accesses (Iterators and Enumerations keep their own state) as 
> its javadoc specifies:
> 
> http://java.sun.com/j2se/1.3/docs/api/java/util/ArrayList.html
> 
> "Note that this implementation is not synchronized. If 
> multiple threads access an ArrayList instance concurrently, 
> and at least one of the threads modifies the list 
> structurally, it must be synchronized externally. (A 
> structural modification is any operation that adds or deletes 
> one or more elements, or explicitly resizes the backing 
> array; merely setting the value of an element is not a 
> structural modification.) "
> 
> 
> Then I wonder whether JDom beta 8 or beta 9, would have 
> problems with concurrent read accesses. I've haven't yet read 
> the code in details, but I think I read somewhere that JDom 
> was internally using lazy initialization when traversing the 
> tree and that concurrent accesses to it might cause problems. 
> Is this [still] true?
> 
> 
> If this turns out that it is unsafe to read/traverse in 
> concurrence the same JDom document, then I would like the 
> group opinion on the best way to implement this cache while 
> avoiding creating a contention point at the JDom document 
> read access: my system is supposed to scale as more computing 
> resources is added (i.e. more CPU in // on the same 
> multiprocessor machine)
> 
> I'm thinking of maintaining a pool of JDom instances. Each 
> concurrent thread would take an instance before traversing 
> it. The multiples instances of the same JDom tree could be created by:
> 1- reparsing the same source
> 2- deep cloning the JDom document
> 3- serializing/unserializing the Jdom Document
> 
> 
> Side question: my document cache needs to be bound in terms 
> of memory usage. I read some threads concerning this in the 
> list, but again do anyone have figures on the amount of bytes 
> used by JDom for storing a parsed representation of a XML 
> stream of N bytes? The experiment I plan on doing is to 
> instanciate M Document instances and look in a profiler at 
> the consummed space once the GC is triggered. Did anybody ran 
> this test before? I did read some data at 
> http://www.sosnoski.com/opensrc/xmlbench/index> .html but this 
> does not quite answers this question.
> 
> 
> 
> Thanks in advance for your help,
> 
> Guillaume.
> 
> 
> _______________________________________________
> To control your jdom-interest membership: 
> http://lists.denveronline.net/mailman/options/jdom-interest/yo
uraddr at yourhost.com




More information about the jdom-interest mailing list