Eric Grinstein <eric@aluno.puc-rio.br> wrote:
> I'm very excited about contributing to Postgres, and thought GSoC
> would be a very good program to get me started.
Great!
> Some commercial RDBMS such as oracle implement a feature called
> 'Query Rewrite' using materialized views. When a query is being
> processed, the system checks whether a materialized view created on
> a similar query could be useful for the query. If so, it rewrites
> it so it can take advantage of the MV, usually dispensing a lot of
> calculation.
Oracle, MS SQL Server, Sybase ASE, and IBM DB2 all have this.
(There may be others.) In essence they treat an MV a bit like an
index, as something you can create to speed up an existing query
without rewriting it. It would certainly be nice to have this in
PostgreSQL, too, in my opinion.
> So, could you give me your opinion about it? If I made a proposal,
> would the community's response be positive? If you have things to
> add, new ideas, I'd be very excited to hear them.
That seems extraordinarily difficult for a GSoC project. Unless
you can demonstrate mastery of the concepts involved in such
optimizations, and a familiarity with the PostgreSQL planner, with
a plan to put forward for how you would do this I think you should
set a more modest goal. Perhaps you could find something to work
on related to the planner that is small enough to be achieved in
the limited time allowed for a GSoC project, that would move you
closer to taking on something this big.
--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
