From: Mark Kremer <mark@...>
Date: 2011-04-08T14:07:27+09:00
Subject: Re: REXML Speed Question

For larger XML documents SAX parsing can really improve performance 
(specifically because SAX parsing doesn't create an entire DOM 
structure, it only extracts the bits you are interested in). Programming 
with a SAX parser is very different though :)

You can also switch to another library for handling your XML, the most 
popular library (at least to my knowledge) is Nokogiri 
(http://nokogiri.org/) and it is a great deal faster than REXML

On 8-4-2011 6:00, Kyle X. wrote:
> Hello, I have been using REXML to extract information from an XML file
> and I am having an issue with the amount of time it is taking.  If I
> point directly to what I want it is pretty fast.  The issue arises when
> I have to grab a reference id, then research for that id to get another
> id, until I finally get to the piece of information I want.  This is
> what a snippet my code currently looks like:
>
> ---------------------------------------------
> result = []
> wall_refs1 = XPath.match( $doc,
> "doc:iso_10303_28/uos/IfcWallStandardCase//*[@pos='1']" )
>
> wall_refs1 = grab_id(wall_refs1,'ref')
> #grab_id simply puts the ref's id and puts them into an array
> #output from this would be [["i1741"]]
>
> wall_ref2 = []
> wall_refs1.each do |ref|
>    x =
> REXML::XPath.first($doc,"//*[@id='#{ref}']//IfcExtrudedAreaSolid").attribute("ref").value
>    wall_ref2<<  x
> end
> #Output [["i1738"]]
>
> wall_depth = []
> wall_ref2.each do |ref|
>    x = REXML::XPath.match($doc,"//*[@id='#{ref}']//Depth").map {|element|
> element.text}
>    wall_depth<<  x
> end
> #Output [["120."]]
>
> wall_depth_final = wall_depth.map do |arr|
>    arr.map do |arr2|
>      #this is simply converting to float and rounding to 2 decimles
>      arr2.to_f.round_to(2)
>    end
> end
>
> wall_depth_final
> #Output [["120.00"]
> -----------------------------------------
>
> The problem with doing this is that it takes substantial time for the
> computer to run this, doing this for say 200 elements can take 25
> minutes (I would be guessing the reason it takes so long to run is
> because as some of the xml files are 10,000+ lines and I image it takes
> a while to comb through that).  I have to start from the first location
> and work my way to the final one, and simply cannot run a search to grab
> //depth unfortunately.
>
> Is there a quicker way of accomplishing the same thing, or is time
> always going to be a burden?
>
> Thank you for your time.
>
> This would be the xml I am reading:
>
> <IfcWallStandardCase id="i1677">
>    <Representation>
>      <IfcProductDefinitionShape id="i1747">
>        <Representations id="i1750" exp:cType="list">
>          <IfcShapeRepresentation exp:pos="0" xsi:nil="true" ref="i1708"/>
>          <IfcShapeRepresentation exp:pos="1" xsi:nil="true" ref="i1741"/>
>        </Representations>
>     </IfcProductDefinitionShape>
>    </Representation>
> </IfcWallStandardCase>
> <IfcShapeRepresentation id="i1741">
>    <Items id="i1746" exp:cType="set">
>      <IfcExtrudedAreaSolid exp:pos="0" xsi:nil="true" ref="i1738"/>
>    </Items>
> </IfcShapeRepresentation>
> <IfcExtrudedAreaSolid id="i1738">
>    <Depth>120.</Depth>
> </IfcExtrudedAreaSolid>
>