Hello Raul,
I have build CVS version and it works fine. My results are: using 1.1: 901ms using cvs: 320ms
Three times faster than 1.1 and three times faster than IBM XML Security Suite according to my tests.
Thanks Raul!!!
Regards,
Egor Pervuninski
egor {dot} pervuninski {at} gmail {dot} comRaul Benito wrote:
At least, fixed!!!!!!!!!!!!!!!!! It was driving me mad. I was thinking of saying: OK, I have do test with hundreds of XML files, and the 1.2 approach was speedier than the 1.1 approach, but your test is clear and we need to throw all the 1.2 version and begin with the 1.1 and tar myself, and wear a paper bag over the head all my life.
But, I'd found the error after using jrat profiler: your test was only spending 6% of the time in the xml-sec code, the rest was spent somewhere else, and that place was toByteArray() inside ByteArrayOutputStream. So with this date I made my simple OwnByteArrayOutputStream, and no matter how big I made the byte array I always got an OutOfBounds exception. Then I remember that I made some changes in order to reduce memory consumption that avoids the reuse of canocilizer as it does not reset the ByteArrayOS. I'll fix this problem(one line change) and I got these new figures:
Your test case, in my machine: using 1.1: 1116 ms using CVS HEAD: 341 ms
A nice improvement, isn't it?
You can test your self obtaining & compiling the CVS HEAD version library or just changing your test to not reuse Canonicalizer object.
Regards,
Raul http://r-bg.com
p.s. Before finding the error, I was thinking that the
NameSpaceSymbTable was the main culprit of the slow down. And I was
thinking in rewriting it in order to not use HashMap and use a simple
String[], it will improve the things if there are few concurrent
different prefix mappings.
Has anybody found more than 10 xmlns:prefix mappings in a document?????? I haven't, but I don't know. So please share your knoledge.
On Tue, 18 Jan 2005 11:14:15 +0500, Egor Pervuninski <[EMAIL PROTECTED]> wrote:
Hello Raul,
I am using Microsoft Windows 2000 Professional 5.00.2195 with ServicePack 4 and "Russian" locale. I have Celeron 1200MHz machine with 512Mb RAM.
I have tried to run java with -server option using two different JREs but results were the same.
Here is the output of java -version for j2sdk 1.5.0: java version "1.5.0" Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0-b64) Java HotSpot(TM) Client VM (build 1.5.0-b64, mixed mode)
Results with -client: 1.1: time spent 921 ms 1.2: time spent 27710 ms Results with -server: 1.1: time spent 2374 ms 1.2: time spent 26438 ms
Here is the output of java -version for j2sdk 1.4.2: java version "1.4.2_03" Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_03-b02) Java HotSpot(TM) Client VM (build 1.4.2_03-b02, mixed mode)
Results with -client: 1.1: time spent 901 ms 1.2: time spent 23014 ms Results with -server: 1.1: time spent 2013 ms 1.2: time spent 23014 ms
I will try to profile the library in my spare time and post the result.
Regards, Egor Pervuninski egor {dot} pervuninski {at} gmail {dot} com
Raul Benito wrote:
Hi Egor,
First of all sorry for not answering your request sooner. Regarding your problem, it seems that the way we transform the stream to UTF8 in 1.2(the library do the conversion itself) is slower in your case to the way it was done 1.1(we use classes given by the SDK). This is weird because, I made a lot of test regarding it and was always faster doing in the library that letting the work for the API. I will investigate this a little more, but IŽll need more information regarding your runtime: What JRE are you using, in what OS, in what locale, and of course are you testing with JIT enable(i.e. are you using -server or -client).
Regards,
Raul http://r-bg.com
On Tue, 11 Jan 2005 18:38:40 +0500, Egor Pervuninski <[EMAIL PROTECTED]> wrote:
Hello,
Sure, here it is.
## Davanum Srinivas : Tue, 11 Jan 2005 08:28:24 -0500
DS> Can you please post your test.xml as well? thanks, dims DS> DS> On Tue, 11 Jan 2005 18:31:18 +0500, Egor Pervuninski DS> <[EMAIL PROTECTED]> wrote:
Hello,
Recently I tried to upgrade my project to use new Apache XML Security 1.2 for Java and found that now the library does c14n much MUCH more slowly than it was with 1.1.
Here is a simple test:
import javax.xml.parsers.DocumentBuilder; import javax.xml.parsers.DocumentBuilderFactory;
import org.apache.commons.lang.time.StopWatch; import org.apache.xml.security.c14n.Canonicalizer; import org.w3c.dom.Document; import org.xml.sax.InputSource;
public class Test { private final static int REPEAT = 1000;
private StopWatch timer = new StopWatch(); private Canonicalizer canon = Canonicalizer.getInstance(Canonicalizer.ALGO_ID_C14N_OMIT_COMMENTS); private Document document;
public Test(String file) throws Exception { DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance(); factory.setNamespaceAware(true); factory.setValidating(false); DocumentBuilder builder = factory.newDocumentBuilder(); document = builder.parse(new InputSource(file)); }
public void runTest() throws Exception { timer.start(); for (int i = 0; i < REPEAT; i++) { canon.canonicalizeSubtree(document); } timer.stop(); System.out.println("time spent " + String.valueOf(timer.getTime()) + " ms"); }
public static void main(String[] args) throws Exception { org.apache.xml.security.Init.init(); Test test = new Test(args[0]); test.runTest(); } }
Just drop commons-lang-2.0.jar, commons-logging.jar, xercesImpl.jar, xml-apis.jar and xmlsec.jarto the classpath and run java Test File.xml.
Version 1.1 output: time spent 952 ms Version 1.2 output: time spent 29172 ms
Where is the catch?
Regards, Egor Pervuninski egor {dot} pervuninski {at} gmail {dot} com
DS> -- DS> Davanum Srinivas - http://webservices.apache.org/~dims/ DS>
Regards, Egor Pervuninski egor {dot} pervuninski {at} gmail {dot} com
