I'll be posting the code to github in the next few days. The value proposition I was looking at was: better performance that SDB ability to use a standard well supported database. I was looking for things like live backup and such.
The bloom filters are used to reduce the number of triples that need to be checked for query, insert, etc. So the final check is still does the filter identified by the bloom filter actually match the triples in the parameters. I should write up a design document as well. Basically, Nodes are not stored directly (i.e. there is not a "node" table) Triples are stored with their bloom filter (and some overhead to make searching more efficient) Triples are stored on pages. Pages have bloom filters identifying the triples on the page. multiple pages comprise the database. I added MySQL bloom filter operations to perform bloom filter checks in the DB and not move the data across the network unnecessarily. I'll let you know when I have the code up. Claude On Mon, Mar 9, 2015 at 10:33 AM, Rob Vesse <[email protected]> wrote: > I'd be interested to see it > > Bloom filters are a useful data structure for some scenarios though I'd be > concerned that a graph backed by a bloom filter is subject to data loss > because it would be possible to not insert some triples because the filter > thinks they are already present as you can always get false positives. > > Also I wonder what value such an implementation has because you still have > to store the triples somewhere so does it actually save any memory? I.e. > what is the value proposition of such an implementation > > Rob > > On 08/03/2015 18:34, "Claude Warren" <[email protected]> wrote: > > >Hello. > > > >I have recently completed a bloom filter based graph implementation that > >works both in memory and with MySQL. I think that the MySQL > >implementation > >is more performant that the SDB implementation, though I have not > >completed > >testing. > > > >The MySQL implementation requires a UDF to handle the Bloom searching. > > > >I am wondering if anyone is interested in this implementation. > > > >Claude > > > >-- > >I like: Like Like - The likeliest place on the web > ><http://like-like.xenei.com> > >LinkedIn: http://www.linkedin.com/in/claudewarren > > > > > -- I like: Like Like - The likeliest place on the web <http://like-like.xenei.com> LinkedIn: http://www.linkedin.com/in/claudewarren
