Hello Michael, thanks for your friendly and fast response.
Concerning the choice of a list and a hashtable depending on the enum count, I did have the same idea. This separation figure depends on the performance of the hashCode() method compared with the equals() method. We had a similar decision in some earlier project. Do you know, which version of Xerces is included in the JDK 5.0, is this the Xerces 250? Would you like me to switch to the [email protected] list, or discuss this issue here? Best regards Juergen Pill > --- Ursprüngliche Nachricht --- > Von: Michael Glavassevich <[EMAIL PROTECTED]> > An: [email protected] > Betreff: Re: Xerces 250: enumeration performance problem > Datum: Thu, 5 Jan 2006 02:35:28 -0500 > > [EMAIL PROTECTED] wrote on 01/04/2006 08:26:18 AM: > > > Dear Xerces committers, > > > > I was evaluating the performance of the XML Schema enumeration language > > element and did detect that it does not scale very well, when the number > of > > enumeration data is scaled within a single enumeration. All other tested > > language elements do scale with the size of the input file (the xml > > instance). > > <snip/> > > > The reason for this N*N complexity lies in following method (Xerces > 250): > > > > package org.apache.xerces.impl.dv.xs; > > public class XSSimpleTypeDecl implements XSSimpleType { > > private void checkFacets(ValidatedInfo validatedInfo) throws > > InvalidDatatypeValueException { > > > > .... > > //enumeration > > if ( ((fFacetsDefined & FACET_ENUMERATION) != 0 ) ) { > > boolean present = false; > > for (int i = 0; i < fEnumeration.size(); i++) { > > if (fEnumeration.elementAt(i).equals(ob)) { > > present = true; > > break; > > } > > } > > if(!present){ > > throw new > > InvalidDatatypeValueException("cvc-enumeration-valid", > > new Object [] > > {content, fEnumeration.toString()}); > > } > > } > > > > ..... > > > > > > The for-loop is executed N/2 times in average, if N is the number of > > enumeration data elements. This for-loop causes the non scaling > performance. > > The code is quite different in Xerces 2.7.1 but it's still doing a linear > search. > > > I would like to suggest to re-implement this part using a hastable > instead > > of a linear list. Would some committer be interested in implementing > this > > change? > > For small values of N the hashtable might have poorer performance compared > to searching a list. I'd favor having a choice between a list and a > hashtable depending on the size of the enumeration. I implemented > something similar for attribute scanning (which wasn't scaling well) a > couple years back. I'll take a look at it. Thanks for bringing this to our > attention. > > > Is this the "right" mailing list for this issue? > > [email protected] might have been a better place to post this but > we're not that picky. :) > > > I can provide more details if requested. > > > > > > Best regards > > > > Juergen Pill > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > Michael Glavassevich > XML Parser Development > IBM Toronto Lab > E-mail: [EMAIL PROTECTED] > E-mail: [EMAIL PROTECTED] > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
