Re: RIM 0.54 and RS 0.83 distribution

[Michael Rowley <mrowley@exceloncorp.com>]

Farrukh,

[resent with CC to the list]

I was reading your query BNF and had some questions.

Are you sure you want the "IN" clause, with arbitrary nesting of
subqueries?  I think this would be especially hard to implement
efficiently with anything other than a relational databse.

In general, it looks like your subset would allow queries that I think
would be difficult to implement if you stored your data in XML, or any
form other than in a relational database:

For example, the following queries for all unique combinations of the
values from the x and y columns of table 'a', and the z column of table
'b'.

SELECT DISTINCT a.x, a.y, b.z FROM a, b

Here is one that would do a 3-way join:

SELECT FROM a, b, c WHERE 
   a.x = b.x and b.y = c.y and a.z = c.z

Of course, nested subqueries could be used to create some hairy requests
as well.

In one of your recent email's, you said that you implemented the query
processor for this BNF in 4 hours, but mentioned that you hadn't bound
it to your back end yet.  It seems like you have primarily just done
parsing then, which is usually easy, so it doesn't surprise me that
it would only take a few hours.

One of the main arguments made after the invention of SQL was that you
can create a language which appears to be asking for very run-time
expensive behavior (joins, cross-products, etc), but that with a
sophisticated query optimizer, it should be possible to actually execute
the query in a reasonable amount of time.  I would hate to require that
people either use a relational database, or implement their own query
optimizer.  It is hard to write a query optimizer.  That is one of the
main reasons that JDO chose a filtering-based query feature, which does
not open the possibility of arbitrary joins and cross-product creation.

Michael