RE: [xml-dev] SAX parser for Netbeans [PERFORMANCE]

> -----Original Message-----
> From: Petr Kuzel [mailto:[email protected]]=20
> Sent: Thursday 20 March 2003 13:55
> To: [email protected]
> Subject: Re: [xml-dev] SAX parser for Netbeans [PERFORMANCE]
>=20
>=20
> Mike Skells wrote:
>=20
> > Hi,
> > Sorry for the cross post but I was not sure which list to post this=20
> > to.
> > =20
> > I presume that NB uses the standard SAX parser available with JDK
>=20
> See org.openide.xml.XMLUtil test:
>=20
>  System.out.println("" + XMLUtil.createXMLReader(false,=20
> false).getClass());  System.out.println("" +=20
> XMLUtil.createXMLReader(false, true).getClass()); =20
> System.out.println("" + XMLUtil.createXMLReader(true,=20
> false).getClass());  System.out.println("" +=20
> XMLUtil.createXMLReader(true, true).getClass());
>=20
> produces:
>=20
>  class org.apache.crimson.parser.XMLReaderImpl
>  class org.apache.crimson.parser.XMLReaderImpl
>  class org.apache.xerces.parsers.SAXParser
>  class org.apache.xerces.parsers.SAXParser
> =20
> It means that it does not use standard JRE parser. It selects=20
> proper one based on performance characteristics we measured in past.
>=20
> > Has anyone looked at using Piccolo http://piccolo.sourceforge.net/=20
> > parser. The tests that I have seen indicate that it is=20
> significantly=20
> > faster.
>=20
> What about conformance?
Conformance info available from
http://piccolo.sourceforge.net/bench.html#summary for the xmlconf
http://xmlconf.sourceforge.net/ conformance suite. The information
presented there indicates the piccolo is _more_ cormormant than crimson
or xerces!
=20

>  =20
> > I dont know how to configure the sax parser for NB, and I=20
> dont have an=20
> > environment that can test the performance, as it needs to=20
> be tested on=20
> > the actual files that are frequently used.
> > =20
> > This could be an easy performance boost for all of NB couldn't it?
>=20
> Possibly, the first start only. Then we turn almost all=20
> information contained in XML into binary format caches.

SO this means that it would provide a faster startup.
Also every time I open a new directory the .nbattrs file is parsed.
Every time I open a form the .form file is processed


For backgroug information I attach an email from the dom4j newsgroup
which has found a 35% gain using Piccolo over Xerces 2.3.0.

There are a series of benchmarks available on
http://piccolo.sourceforge.net/bench.html#benchmarks which compare
crimson xerces picolo and others



--forwarded message --



<DISCLAIMER>
The test was NOT any official test.  All trademarks used in this email=20
are those of their respective organizations.  This email contains no=20
warranty expressed or implied and is "AS IS" information.  This is not=20
for any official useage.  By reading this you release Lucas Software=20
Engineering, Inc. and/or David D. Lucas of any liability in relationship

to this information.
</DISCLAIMER>

I don't see anything wrong with sharing the information, as long as no=20
one is on a benchmark hunt.

Some informatino about the simple test.  It was an XML parse using some=20
software that I am productizing.  The way I use the DOM4J processing is=20
the same regardless of SAX parser.  The numbers are not a raw indicator,

but a relative indicator with Xerces 2.3.0 as a baseline. Numbers=20
greater than 1.0 represents a faster parse.  The tests were run on an 2=20
CPU (1MHz) SMP Linux machine running Sun JDK 1.4.1.  My future plans are

to package piccolo as part of my releases as a default parser.

Another reason I now like piccolo is the fact that various versions of=20
Xerces are all over the place.  Particularly inside application servers=20
you end up having to use the one that the base classloader gives you.

I hope this information is helpful.  :-)

Later,
Dave

Mike Skells wrote:
> David,
> I would like to forward this email to the netbeans dev newsgroup=20
> unless you have any objections.
>=20
> Within netbeans there are a large number of XML document parsed=20
> regularly, so this should benefit that community.
>=20
> Mike
>=20
>=20
>>-----Original Message-----
>>From: [email protected]
>>[mailto:[email protected]] On Behalf Of=20
>>David D. Lucas
>>Sent: Saturday 22 February 2003 03:37
>>To: David D. Lucas
>>Cc: Carlos Barroso; [email protected]
>>Subject: Re: [dom4j-user] Has someone tested this?
>>
>>
>>My simple DOM4J test REVISED to include Piccolo 1.03, WOW!
>>I see a 35% gain using Piccolo over Xerces 2.3.0.  Hmmm, I may end up
>>packaging Piccolo with this boost. :-)
>>
>>
>>Parser                   |      Gain
>>=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D+=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>>Xerces 2.0.1             |      0.98
>>Xerces 2.3.0             |      1.00
>>FastParser (JAPISoft)    |      1.16
>>JDK 1.4.1 (JAXP default) |      1.18
>>
>>Piccolo 1.03             |      1.35
>>


--=20

+------------------------------------------------------------+
| David Lucas                        mailto:[email protected]  |
| Lucas Software Engineering, Inc.   (740) 964-6248 Voice    |
| Unix,Java,C++,CORBA,XML,EJB        (614) 668-4020 Mobile   |
| Middleware,Frameworks              (888) 866-4728 Fax/Msg  |
+------------------------------------------------------------+
| GPS Location:  40.0150 deg Lat,  -82.6378 deg Long         |
| IMHC: "Jesus Christ is the way, the truth, and the life."  |
| IMHC: "I know where I am; I know where I'm going."    <><  |
+------------------------------------------------------------+

Notes: PGP Key Block=3Dhttp://www.lse.com/~ddlucas/pgpblock.txt
IMHO=3D"in my humble opinion" IMHC=3D"in my humble conviction"
All trademarks above are those of their respective owners.