Datalog notes

Why Datalog over SPARQL?

• Datalog allows general recursion; SPARQL only allows structural recursion in recent version
• Lots of research on optimization in the 80's
• Some commerical success (http://www.logicblox.com/)

use cases

• applications that generate queries
• alerts
• people who write queries

example systems

• client database
  • pull-down menu-ish thing
  • originally sgi uv, porting to xmt2
• lexis-nexis

Other semantic databases

• http://en.wikipedia.org/wiki/Virtuoso_Universal_Server

{{{ blue-modern alice->bob: Test bob->alice: Test response }}}

example from insurance people

• triple: { id, name, address }
• id is unique
• id maps to multiple names
• name maps to multiple addresses
• find all ids that share two or more addresses with another id
• or that share single address, compute distance function on name and select if close enough

relational algebra -> nested loops or whatever

• select
• project
• join
• others
  • aggregation
  • union
  • application-specific stuff

stuff

• greenplum (has some column-oriented stuff but not fully
• virtuoso (13-hour tournament query)
• SPARQL benchmarks? (http://blog.dydra.com/2011/05/27/sp2b-benchmarks)

john's problem set

• word problems

• c loops

• sp^2bench : a sparql benchmark suite (icde 09)

  • q7 is the one that's hard.

• relational algebra (database textbook: )