RE: Really, Really fast validation

11 Dec 2001 00:03:58 -0000

Hi Vlad,
 
That is the case, the JVM is being started only once. We are not starting it every time we try to parse a message. Actually, the code being used on solaris is the same as that being used on NT. The only difference is a dll version. We have an NT version and a solaris version, both of which should have the same C code.
 
 
Thanks 
 
Holger
 
 
 
 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Holger Paffrath
Technical Consultant
 
Suite 119, 566 St Kilda Rd
Melbourne 3004
Tel: +61 3 9533 9999       
Fax: +61 3 9533 9988
Mob: +61 418 611 345     
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 
 
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Tuesday, December 11, 2001 2:37 AM
To: xerces-j-user
Subject: RE: Really, Really fast validation


Hello Holger,

In my case Solaris has always run faster then PC, but I use servlets to parse and process XML messages.
You just have to make sure that JVM stays up all the time.
It takes quite a bit of time to start JVM and load up all the Xerces classes even to parse a small XML document.
Using of EntityResolver for caching purposes will not make any difference if JVM is restarted for every request.
VAJ and most of other development tools keep JVM running as long as development environment is up and if you are using one of the dev. tools - it could be the reason why your PC parses XML faster than Solaris machine.

Regards,
Vlad


Please respond to xerces-j-user <[EMAIL PROTECTED]>

To:        xerces-j-user <[EMAIL PROTECTED]>
cc:        

Subject:        RE: Really, Really fast validation

I've found the problem, I just don't know how to fix it.
 
The problem we have found is that our xerces parser is being called through a C program which instantiates a JVM.
There is a known problem (or so I'm told) that streams run slower under solaris (The platform we are using). What we are finding is a 10x increase in processing time.
The problem doesn't occur on the windows platforms.
With a process that is time dependant, this is a very big problem.
 
Using timestamps in my program. I've determined that the problem is within xerces parse method.
 
I suspect that the entityResolver uses a stream which runs into the above problem causing performance problems.
 
Does anyone have a method of specifying an .xsd to validate a message against without using a stream to buffer the message?.
 
I've tried using the property external-schemalocation, but it takes even longer (15x increase).
 
 
 
Anyone got any ideas?
 
Thanks
 
Holger
 
 
 
 
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Tuesday, December 04, 2001 11:59 PM
To: xerces-j-user; holger
Subject: Re: Really, Really fast validation


I've been trying to get that working, but unfortunatelly Xerces1 doesn't have interface to implement
schema grammar caching (1), which means that for every message it parses your schema file(s) over again.
Maybe somebody else can comment on whether schema grammar caching can be implemented with Xerces2..

1. http://xml.apache.org/xerces-j/faq-general.html#faq-5

Please respond to xerces-j-user <[EMAIL PROTECTED]>; Please respond to holger <[EMAIL PROTECTED]>

To:        "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
cc:        

Subject:        Really, Really fast validation

Does anyone have any tips or tricks in validating an XML schema really, really fast. Ie. sub second validation with 1-2k messages and 370k of schema files.


I've tried loading the schema to a hashmap, but it still takes 2-10 seconds to validate depending on the platform (NT and Solaris) and java version (1.1, 1.3).

Thanks

Holger




--------------------------------------------------------------------
mail2web - Check your email from the web at
http://mail2web.com/ .


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to