From: Danny Wyatt (danny@cs.washington.edu)
Date: Wed May 19 2004 - 01:43:43 PDT
This paper provides a good, textbook-level overview of techniques
(perhaps not exactly "textbook-level": I did turn to Ramakrishnan for a
few lengthier explanations) for executing operations on sets and
relations. It was interesting to see topics from "data structures and
algorithms 101" put into practice with anecdotal evaluation results from
some real systems. I have a better understanding of the trade-offs
involved.
On the topic of planning and execution of physical queries from logical
queries, I am curious whether any logical query (or class of logical
queries) can always be optimized at the physical level (albeit to the
detriment of other classes of logical queries) or whether there exist
some logical queries that are ineluctably hard to execute no matter the
physical representation, plan, and algorithm.
As has been pointed out, this paper is a decade old (it's brief concern
with object-oriented databases betrays that). He points out that
methods based on hashing were previously unexamined but have become
generally accepted at the time of his writing. I assume no other broad
developments such as that have occurred in the meantime. I do wonder
what the status of certain "new" areas he presents is
today---particularly the method for dynamically selecting query subplans
based on statistics gathered during the run of earlier subplans. I'm
also curious how research into semi-structured querying has affected
algorithms for nested relations.
Lastly, as a minor historical question, do we owe the now ubiquitous
"iterator pattern" to the databases community?
This archive was generated by hypermail 2.1.6 : Wed May 19 2004 - 01:43:48 PDT