Another area is possibility to build spatial index - it should make map paint and functions like getNearestNode much faster. Spatial index and painting only primitives that are not too small is imho the main reason why is josmng so much faster than josm.
Btw. Currently I'm not continuing on the refactoring until JOSM becomes a bit more stable again and new JOSM tested is released. -- jttt On Tue, Jul 28, 2009 at 8:24 PM, Karl Guggisberg<[email protected]> wrote: > I saw it too and I hope it's going in the direction of > > First > ** Encapsulation > - getting rid of most if not all public field in Node, Way, Relation, and > DataSet > - providing a richer API for compound primitives like ways and relations, > i.e. > Relation::getReferringPrimitives() -> Set<OsmPrimtive> > Way::isConnectedTo(Way other) -> boolean (or an enumeration > "undirected", "headToTail", "headTohead", etc.) > Relation::getMemberReferringTo(OsmPrimitive primitive) -> RelationMember > Relation::refersTo(OsmPrimtive primitive) -> boolean > etc. etc. > - providing a cleaner state model for OsmPrimitives (including unmodified, > incomplete, modified, deleted, visible) > - possibly getting rid of the "selected" flag in primitives and adding > selection management to > DataSet, including emitting selection change events > (no more DataSet.fireSelectionChanged(...) after every update of the > selection) > > Then > ** Efficiency > - possibly optimizing the then encapsulated datastructures of Dataset, i.e. > in order to optimize backreference lookup, > as it is already probosed in the backreference visitor. Code for deleting > primitives, for "purging" primitives, and > for parsing datasets would benefit from such an improvement too > - perhaps we can avoid to copy around large sets of primitives as the > allXYZ-Methods on DataSet do it today > > Additionally: a general cleanup of the code in the core data classes > wouldn't be no harm either > > -- Karl > > > -----Ursprüngliche Nachricht----- > Von: [email protected] > [mailto:[email protected]] Im Auftrag von Dave Hansen > Gesendet: Dienstag, 28. Juli 2009 19:21 > An: josm-dev > Betreff: [josm-dev] jttt and private class members > > I've seen some SVN commits from jttt doing some refactoring of the > Way.nodes. First of all, very cool stuff. I'd like to use this to make > some of my ReverseLookup code and a bunch of the validator tests faster. > > What's the end goal of all of the churn, though? > > -- Dave > > > _______________________________________________ > josm-dev mailing list > [email protected] > http://lists.openstreetmap.org/listinfo/josm-dev > > > _______________________________________________ > josm-dev mailing list > [email protected] > http://lists.openstreetmap.org/listinfo/josm-dev > _______________________________________________ josm-dev mailing list [email protected] http://lists.openstreetmap.org/listinfo/josm-dev
