Implmentation Optimizations - lambda-land/VDBMS GitHub Wiki
These are possible optimizations we can do to improve the runtime of vqueries besides applying the variational relational algebra optimization rules to the vqueries before running them:
- Incremental query evaluation by using temp tables. e.g. run
q1 join q2
only once inq1 join q2 join f<q3,q4>
. - Considering operations on tuples' presence conditions as operations on strings instead of converting them to feature expression back and forth.
- Smart union before join?
q1 join q2 join f<q3,q4>
it might be more performent to disjoin the result ofq3
andq4
before joining them withq1 join q2
? Ask Eric! - Implement feature expressions as UDT and their operations as UDFs in the database engine.
- Extend the SqlValue library to allow for adding a data constructor for feature expressions and their functions?!