From: Robert Klemme <bob.news@...>
Date: 2005-08-31T17:21:23+09:00
Subject: Re: SAX-based XPath

Gary Shea wrote:
> I'm posting this in hope of getting some API suggestions.
>
> I'm building a native stream-based Ruby XPath processor (or whatever
> it would be called) in order to parse some gigabyte-scale XML files at
> work.  It will accept multiple XPath expressions and output events
> (SAX for now) matching the union of the XPath expressions.
>
> It currently only works with absolute, non-wildcarded, predicate-less
> default-axis XPath expressions:
>
> filter = XmlFilter::XPathFilter.new
> filter.listener = XmlFilter::RecordingListener.new
> filter.xpath = '/a/b/c'
>
> parser =
> REXML::Parsers::SAX2Parser.new(File.open('some_file_path.xml'))
> parser.listen = filter
> parser.parse
>
> This interface needs to be extended a little to work with multiple
> XPath expressions, maybe:
>
> filter.xpath = ['/a/b/c', '/d/e/f']
>
> Any suggestions for a more Ruby-esque way to do it?

It seems you could simplify the interface a bit (or add a method) along
the lines of REXML so you can do

File.open('some_file_path.xml') do |io|
  XmlFilter::XPathFilter.parse(io, '/a/b/c', '/d/e/f') do |event, filter|
    # process event
  end
end

I'm unsure about the "filter" block parameter but it might be useful to
know the matching filter criterium.  What do you think?

Kind regards

    robert