Query Evaluation

Fundamentals

SQL is Complex - Reduce to "Relational Algebra" 
Deal with Nested Queries one block at a time 
Issue of Correlation Variables 
For each block build up a query tree with operators at nodes 
Most operators are unary, joins are binary 
Operators of greatest interest are selections, projections, and joins 
"Sort" may also be an "operator" - the form of the data is important 
Notion of "Physical Algebra" 


Pipelining

Intermediate Results can be large - must count cost for writing/reading 
Avoid writing intermediate results if possible - pipelining 
Structure Code as recursive calls, top-down. 
Each operator supports open, close and get_next 
For scan over relation, gives rise to concept of cursor 
"Blocking Operators", e.g. sorting, may do all their work on the open 


Internet/XML Query Processing

Structured Join --  Join based on position rather than value
Non-blocking join -- required when inputs may take a long time to complete