To jump in here, the DataStore API was developed specifically to provide live connections to DBs. The DataSource API was old code that was just our first quick attempt at abstracting the concept of connecting to data sources. It was basically obsoleted by the DataStore work.
We also found threading problems with making an ArcSDE DataStore interface (which BTW we completed in-house at Vivid, but never released). The problem is that the SDE API is inherently single-threaded, but not synchronized. The solution was to wrapper the SDE API completely and force it to be synchronized on a single thread. I can provide more details if you're interested. I would strongly recommend extending the DataStore design. It was intended to be able to support R/W as well, although we never got to the point of implementing this. For a model of building R/W drivers, have a look at GeoTools - they solved this well, I think. I don't think it's too much of a mystery, but some thought needs to be given to how to identify keys for records. Larry Becker wrote: > Hi Paolo, > > Thanks for your thoughts. I'll defer to your experience in > implementing database plugins since I wasn't directly involved with > coding the ArcSDE plugin. The problem I have with the DataStore > ArcSDE driver is that it is inherently multithreaded when you have > more than one layer, so it seems to bring out the worst in both the > JUMP code and in the database system, especially Oracle, which runs > out of database connections much too easily. > > I agree with your statement about the caching mechanism. I think > the class you are referring to is: > com.vividsolutions.jump.workbench.model.cache.CachingFeatureCollection. > I like your idea of adding a "dirty" cache to hold edited records. > I'll give some thought to how it might be done. > > I do have a little experience with using H2 on another project. I > can report that I found it to be much faster in every way than Derby, > especially the embedded H2. In general, it was a joy to work with. > It has an active maintainer and good online documentation. > > regards, > Larry > > On Thu, Apr 23, 2009 at 4:27 AM, Paolo Rizzi <g...@oicom.com > <mailto:g...@oicom.com>> wrote: > > > Thanks Stefan. Also, as far as the best database base class to > > consider, we have DataSource which loads the layer into memory, and > > DataStore which doesn't but is read-only, and IMHO has problems with > > multiple layers due to threading. I guess, if I was starting > the task > > of implementing a new database driver, I would start with a > DataSource > > driver. Then once I had that working I would implement a DataStore > > driver. Then I would consider how best to fix the DataStore > model so > > that it is more reliable and supports read/write. There is also the > > option of building a new database base class, but it would > require a lot > > of core mods to integrate into File->Open, etc. > > > > Larry > Just my little 2c (as always I'm sorry of not contributing any > code...). > > Regarding the summer work of H2 integration, if I recall well, > Christopher talked about: "...very large TINs..." and that seems not > well suited to be read in-memory. > When I wrote my SISDB PostGIS/Oracle plugin, back in 2006, I looked at > the code for both DataSource and DataStore and found that the > DataStore > design was _much_ more clean and carefully thought out than > DataSource. > So I decided to follow the DataStore design and never regretted it, > infact the plugin always worked like a charme, even on machines > with low > memory available. > I never had problems with multiple layers, but it's maybe because > I had > to use a separate reading thread inside the plugin... > > Reagarding the addition of writing capabilities to the DataStore > interface, what is actually needed is something inside the core > itself. > I never studied the code, but I believe there's a "caching" > mechanism of > some sort already in place in OJ, and I think it works upon the > current > view, so if it needs data outside of it, the cache requests it to the > DataStore. I also imagine that upon some condition (memory occupation) > this cache simply discards out-of-view features. > > To implement writing, it is necessary that dirty features > (modified/inserted/deleted) are not discarded, but are kept or > copied to > a "dirty" cache. When the user eventually decide to commit, the dirty > features are to be passed back to the DataStore along with their > corresponding operation (Modify/Insert/Delete). > > To add writing it is not required to modify the DataStore interface at > all, so that existing plugin will continue to work. > A new DataStoreWriter interface may be created for this purpose and an > "instanceof", or something like that, can be used to dinamically > cast a > plugin to be a writer, only if it actually implements that > interface. Or > even a DataStoreWriterProvider interface maybe defined with a single > method DataStoreWriter getWriter(), so that delegation can be used. > > Obviously DB tables are required to have a Primary Key to be editable > and also some sort of conflict resolution mechanism, or at least > something to inform the user of what has happened, but these two > requirements both exists even if following the DataSource paradigm. > > Bye > Paolo > > > > > > On Wed, Apr 22, 2009 at 2:51 PM, Stefan Steiniger > <sst...@geo.uzh.ch <mailto:sst...@geo.uzh.ch> > > <mailto:sst...@geo.uzh.ch <mailto:sst...@geo.uzh.ch>>> wrote: > > > > +1 > > > > I was going to ask by the end of the week if Landon tested > it and if we > > can add the method. Basically I just trust you on that > (having way more > > developer experience than I have). > > > > sorry that databases are not my area if expertise - although > I may know > > some very few things about it > > > > stefan > > > > Larry Becker wrote: > > > To maximize the use of Christopher's time this summer > integrating H2 > > > spatial, we should ensure we have the database code > > infrastructure ready > > > so he can focus on the actual problem. > > > > > > So the first step would be consideration of my proposed > mod to > > > BasicFeature to add an isModified() method. I have been > testing > > this in > > > SkyJUMP and it seems to work as I expected. If you > recall, this > > basic > > > support is needed to make it possible to know what database > > records to > > > update, since (for large dbs) you can't realistically > expect to write > > > the whole database table each time you change a record. > Other > > methods > > > are required to track deletions, and perhaps additions. > > > > > > As far as other database support enhancements go, I would > like to > > hear > > > from Martin. This would seem to be something to which he > might have > > > given some thought. > > > > > > regards, > > > Larry > > > > > > -- > > > http://amusingprogrammer.blogspot.com/ > > > > > > > > > > > > ------------------------------------------------------------------------ > > > > > > > > > > ------------------------------------------------------------------------------ > > > Stay on top of everything new and different, both inside and > > > around Java (TM) technology - register by April 22, and save > > > $200 on the JavaOne (SM) conference, June 2-5, 2009, San > Francisco. > > > 300 plus technical and hands-on sessions. Register today. > > > Use priority code J9JMT32. http://p.sf.net/sfu/p > > > > > > > > > > > > ------------------------------------------------------------------------ > > > > > > _______________________________________________ > > > Jump-pilot-devel mailing list > > > Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net> > > <mailto:Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net>> > > > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > > > > > > ------------------------------------------------------------------------------ > > Stay on top of everything new and different, both inside and > > around Java (TM) technology - register by April 22, and save > > $200 on the JavaOne (SM) conference, June 2-5, 2009, San > Francisco. > > 300 plus technical and hands-on sessions. Register today. > > Use priority code J9JMT32. http://p.sf.net/sfu/p > > _______________________________________________ > > Jump-pilot-devel mailing list > > Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net> > > <mailto:Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net>> > > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > > > > > > > > > > -- > > http://amusingprogrammer.blogspot.com/ > > > > > > > ------------------------------------------------------------------------ > > > > > > ------------------------------------------------------------------------------ > > Stay on top of everything new and different, both inside and > > around Java (TM) technology - register by April 22, and save > > $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. > > 300 plus technical and hands-on sessions. Register today. > > Use priority code J9JMT32. http://p.sf.net/sfu/p > > > > > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > Jump-pilot-devel mailing list > > Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net> > > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > > > > ------------------------------------------------------------------------------ > Stay on top of everything new and different, both inside and > around Java (TM) technology - register by April 22, and save > $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. > 300 plus technical and hands-on sessions. Register today. > Use priority code J9JMT32. http://p.sf.net/sfu/p > _______________________________________________ > Jump-pilot-devel mailing list > Jump-pilot-devel@lists.sourceforge.net > <mailto:Jump-pilot-devel@lists.sourceforge.net> > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > > > > > -- > http://amusingprogrammer.blogspot.com/ > ------------------------------------------------------------------------ > > ------------------------------------------------------------------------------ > Stay on top of everything new and different, both inside and > around Java (TM) technology - register by April 22, and save > $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. > 300 plus technical and hands-on sessions. Register today. > Use priority code J9JMT32. http://p.sf.net/sfu/p > ------------------------------------------------------------------------ > > _______________________________________________ > Jump-pilot-devel mailing list > Jump-pilot-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > -- Martin Davis Senior Technical Architect Refractions Research, Inc. (250) 383-3022 ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensign option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects _______________________________________________ Jump-pilot-devel mailing list Jump-pilot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel
