Performance problem on new WMS Capabilities document parser

[vash]

I've noticed a sensible lack of performance on the WMS capabilities parser, it takes 5 mn to parse my local server capabilities. I've read the code and i think it could be optimized, it do around 1000 xpath request on the document to parse. The biggest problem is that each getter on a field of the document do a xpath.

example :

    public String getVersion()
    {
        return this.getText("*/@wms:version");
    }

a better design could be :

    private String version= null;
    
    public String getVersion()
    {
    	if (version== null) {
    		version= this.getText("*/@wms:version");
    	}
        return version;
    }

I'm refactoring the parser to see if this design improve performance. A sax parser instead of DOM/Xpath would also be a good solution to improve performance.


What do you think about this, Tag?

[vash]

I've finished the redesign. The redesigned parser spent 10s to parse the same file vs 5 mn for the original parser.

http://nicastel.free.fr/worldwind/wms_cap_parser.zip

[tag]

Thanks! What'd you do?

[vash]

Originally Posted by tag

Thanks! What'd you do?

I've applied the pattern explained in my first post on all the methods of Capabalities and derived classes.

For the method which takes an element in parameter such as the Layer Items methods, i've created a inner class which contains all the fields necessary and wrap an element, and i've redesigned the methods associed.

example :

    public String getLayerAbstract(Element layer)
    {
    		return this.getText(element, "wms:Abstract");
    }

becomes

    public class LayerElement {
    	public Element element;
    	
    	public String layerAbstract = null;
    	public String layerAttributionTitle = null;
    	public String layerAttributionURL = null;
    	public String layerAttributionLogoFormat = null;    	
    	public String layerAttributionLogoHeight = null;
    	public String layerAttributionLogoURL = null;
    	public String layerAttributionLogoWidth = null;
    	public MetadataElement[] layerAuthorityURLs = null;
   	    public BoundingBox[] layerBoundingBoxes = null;
   	    public String layerCascaded = null;
   	    public String[] layerCRS = null;
   	    public String layerDataURLFormat = null;
   	    public String layerDataURL = null;
   	    public DimensionElement[] layerDimensions = null;
   	    public DimensionElement[] layerExtents = null;
   	    public BoundingBox layerGeographicBoundingBox = null;
   	    public String layerFeatureListFormat = null;
   	    public String layerFeatureListURL = null;
   	    public String layerFixedHeight = null;
   	    public String layerFixedWidth = null;
   	    public IdentifierElement[] layerIdentifiers = null;
   	    public String[] layerKeywordList = null;
   	    public String layerMaxScaleDenominator = null;
   	    public MetadataElement[] layerMetadataURLs = null;
   	    public String layerMinScaleDenominator = null;
    	public String layerName = null;
    	public String layerNoSubsets = null;
    	public String layerOpaque = null;
    	public String layerQueryable = null;
    	public String[] layerSRS = null;
   	    public StyleElement[] layerStyles = null;
    	public LayerElement[] layerSubLayers = null;
    	public String layerTitle = null;
    	
    	public LayerElement(Element element) {
    		this.element = element;
    	}
    }
    
    public String getLayerAbstract(LayerElement layer)
    {
    	if (layer.layerAbstract == null) {
    		layer.layerAbstract = this.getText(layer.element, "wms:Abstract");
    	}
        return layer.layerAbstract;
    }

With this design only the fields needed by the client are accessed with xpath and only one time for each field.