Re: [Xdoclet-user] XJavadoc is too memory intensive

[Clifton C. Craig]

That was my original intention however it is very, very memory 
intensive. It works pretty good for a small subset of morederately sized 
source but as the source base grows (in our case we have 100s if not 
1000s of source files that must be scanned) it performs very terrible. 
I'd love to resolve this memory issue but I have little time. I have 
been digging through XJavadoc's source and trying to find where the 
problem is. Apparently there is no way to dump SourceClass objects and 
ProgramElement objects after they've been created. Ideally I'd like to 
Create them as needede then dump then for garbage collection when done. 
I noticed a big jump in memory usage when I started scanning for certain 
methods. For example, one part of my tool looks for the ejbCreate method 
by iterating method names. The engine has to create method objects to 
read these names and once done they apparently stick in memory. Ever 
since I put that piece in I get out of memory errors all the time and it 
even slows my WinXP box to a crawl after it errors out. It forced me to 
up the JVM memory to 740M via ANT_OPTS and increase my page file size to 
nearly 2G. Even with this big a page file, a few runs of the build 
leaves me with little or no virtual memory. I'm not sure if this is an 
Ant bug, a JVM bug or what. It seems to me that all the memory (virtual 
or otherwise) should be freed once the process completes. I'd love to 
share the source but I have to check with my boss and see if it breaks 
the confidentiality agreement I signed.

Clifton C. Craig, Software Engineer
Intelligent Computer Systems -  A Division of GBG
2101 Embassy Drive
Lancaster, PA  17603
Phone:  717-295-7977 ext. 621
Fax:  717-295-7683
[EMAIL PROTECTED]
[EMAIL PROTECTED] 



Lee Marlow wrote:

Clifton,

Is this tool something you plan to contribute back to the community?  I know
my team would greatly appreciate it, as we moved to XDoclet halfway through
the project and still have quite a few non-XDoclet EJB's.
Just curious.

-Lee Marlow

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Clifton C.
Craig
Sent: Tuesday, April 13, 2004 7:50 AM
To: [EMAIL PROTECTED]
Subject: [Xdoclet-user] XJavadoc is too memory intensive
Hello all,

I've been toying with using XJavadoc to reverse engineer a bunch of 
deployment descriptors into Java source as XDoclet tags. It's been 
working pretty good except for one thing. The XJavaDoc API sucks memory 
dry. I have a rather large project involving hundreds of EJB java files 
and I cannot find an efficient way to feed them into the XJavaDoc engine 
as source classes when useNodeParser is set to true. I started trying to 
feed them all in as one big SourceSet. Then when I kept getting out of 
memory errors I tried feeding them in 3 at a time creating a new 
XJavaDoc object for each set of 3 EJB beans. I ended up recoding the 
XJavaDocTask and creating a new XJavaDoc object for each class. Now at 
least it works but it runs like a snail. (Do snails run?) I looked for a 
method to empty the source set in the XJavadoc object so I at least 
wouldn't have to null and recreate it each time but I found nothing. 
What's a developer to do?

 



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
xdoclet-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/xdoclet-user