[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Handling very large instance docs

• To: <xml-dev@l...>
• Subject: Re: Handling very large instance docs
• From: "Karl Waclawek" <karl@w...>
• Date: Thu, 29 Apr 2004 10:28:30 -0400
• References: <p06020415bcb6ad5e7762@[89.10.0.19]> <p05200000bcb6b91b64d3@[128.253.109.55]>

> >At the very least I need to be able to sequentially process a large
> >document and extract an identified sub-tree (ideally denoted by an
> >XPath expression) for run-of-the-mill tools to manipulate. I assume
> >such a beast would need to be based on a SAX parser.
> 
> I did exactly that in Python.  I considered building an engine that 
> could filter SAX events to those that match a limited version of 
> XPath, but ran out of gas.  I ended up with a just regular SAX 
> application.

Interesting - I always thought such a thing is useful, but haven't
come across implementation.

I built something like that in Delphi (I call it SAXPath)
on top of SAX. First you define an array of records (structs)
each with a name (or wildcard) - like XPath - and a call-back
interface pointer (used for filtering/predicates or processing).
I call the array elements "path nodes", and the array "path handler".
Then you register such an array with a "handler manager" for processing.
Only relative paths are currently supported.

Call-backs are done on every node of such a "path handler"
as long as it matches and as long as filter-call-backs 
further up haven't de-activated the "path handler".

For the projects I am involved in this has proven very practical.

Karl

• Follow-Ups:
  • Re: Handling very large instance docs
    • From: Kevin Jones <kjouk@y...>

• References:
  • Handling very large instance docs
    • From: Andy Greener <andy@g...>
  • Re: Handling very large instance docs
    • From: Joel Bender <jjb5@c...>

• Prev by Date: Re: Handling very large instance docs
• Next by Date: Namespace arguments, namespace routing language
• Previous by thread: Re: Handling very large instance docs
• Next by thread: Re: Handling very large instance docs
• Index(es):
  • Date
  • Thread