Justin Deoliveira wrote: > Ben Caradoc-Davies wrote: >> Andrea Aime wrote: >>>>> Another option could be to have the build server build with >>>>> -Pextensive (shall we rename this to -Pexpensive ;-) ?) >>>> +1 for the rename. >>> I was just joking :) >> I know, but it was a good suggestion. >> >>> This Eclipse XSD issue seem relevant as well and it's being discussed >>> these days too: >>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=278853 >> Could be relevant. >> >>> Also, an old thread on the eclipse ml suggests to use xerces-j schema >>> api instead of XSD if one does not have schema manipulation needs: >>> http://dev.eclipse.org/newslists/news.eclipse.technology.xsd/msg01221.html >>> >>> This does not mean we have to jump on xerces-j api right away, just >>> that we keep it in mind for the future (this is the same API DeeGree is >>> using for its schema handling needs, too). >> Mmm, we only consume schemas, so xerces-j may be an option. We do need >> to walk the schemas to find things like substitution groups, and I am >> not sure the xerces API allows this. Justin? > > Correct, we do inspect the schema contents quite closely, so any tech > needs to model every part of the schema down to small detail, this > includes particle information, and substitution group info. > > We also do need the ability to build up schemas from scratch, this is > how the application schema stuff works, at least in geoserver. But I > agree, xerces-j is definition worth a look at. > I would hold horses before even thinking on switching xsd libraries, that'd be just too much work. If we let the profiler speak though, it doesn't seem like the biggest offender is EMF XSD itself, but our (my) own code. For example, if I run _only_ the FeatureChainingTest mappings load method through the profiler (that is, parsing the mappins, the xsd schemas to EMF XSD and the EMF XSD model to FeatureType/AttributeDescriptor), it needs 74M Heap space. But if I omit the last step (that is, parsing the mappins and the xsd schemas to EMF XSD, omiting the transformation from EMF model to FeatureTypes/AttribtueDescriptors), it only needs 22M Heap. That is, the geotools model takes ~2.5 as much space as the EMF one, hence my suggestion of transforming only the _needed_ EMF objects to GeoTools ones instead of the whole schemas.
aside, see my previous post "app-schema build time/mem" for some more insight on optimizing build resources. -- Gabriel Roldan OpenGeo - http://opengeo.org Expert service straight from the developers. ------------------------------------------------------------------------------ Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge _______________________________________________ Geotools-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/geotools-devel
