[
https://issues.apache.org/jira/browse/XALANJ-2398?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12529636
]
Michael Glavassevich commented on XALANJ-2398:
----------------------------------------------
Brian, the commit I referenced is the one in which SAX2DOM started using
CharacterData.appendData() to accumulate text. The current version in Apache
calls that method and as I pointed out to Shereef using
CharacterData.appendData() to accumulate text can be very inefficient. A
StringBuffer will do the job much better (and in fact that's we use the DOM
parser in Xerces instead of this method).
> parsing big XMLs take very long time. A JAXRPC webservice request of soap
> size 2MB takes above 5 mins to complete parsing and start processing web
> service.
> -----------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: XALANJ-2398
> URL: https://issues.apache.org/jira/browse/XALANJ-2398
> Project: XalanJ2
> Issue Type: Bug
> Components: transformation
> Environment: Operating System: All
> Platform: All
> Reporter: SHEREEF ABDULLA
> Attachments: JaxRPCProcessRequest.java, WrkingBigSOAP.txt
>
>
> parsing bix XMLs take very long time. A JAXRPC webservice request of soap
> size 2MB takes above 5 mins to complete parsing and start processing web
> service.
> Observed that the SOAP messages are read line by line and
> CharacterDataImpl.append() is getting called for each line. This appends each
> line to the string data which stores the previously read xml part. This call
> results in many string additions (say 20000+ for 1MB SOAP) of big string and
> the thread is most of the time blocked in StringBuilder.expandcapacity due to
> long string additions.
> JAX RPC webservice with SOAP messages bigger than 1MB take 5 mins and more
> just for the web service implementation to start working as the jaxrpc does a
> SOAPMessage.getenvelop call to do HandlerChainImpl.checkMustUnderstand()
> checking for the request message header. Same problem happens for response
> also. for time being we commented checkMustUnderstand method so that the
> parsing doesn't happen at all.
> String additions for each of the lines would have been avoided, either by
> creating the whole data single time or using string buffer instead of string.
> I tried to modify the data field to use StringBuffer instead of String but
> the underlying CoreDocumentImpl.modifiedCharacterData() and all underlying
> calls takes the string params so couldn't go ahead with it.
> bug # XERCESJ-102 looks like the same issue..
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
