From: Arkadiy Vertleyb (vertleyb_at_[hidden])
Date: 2002-04-30 17:42:11
>>it should be possible to separate
>>your algorithms from how the data is stored.
I agree. Right now all the operations are templates that can accept any
type of parameters as long as they provide certain typedefs and implement
certain functions. It just happened that we implemented our table class on
std::vector. We think that concrete table implementations can be done by
the users of the library.
>>Also, some relational operations can be much more efficient if indexes are
>>available, so you should support optional indexing.
This is more difficult question. Right now we don't have any indexing,
although we do have plans to think about it in the future.
However, I think we should be careful not to get into the business of
writing yet another RDBMS. There are plenty of folks around who can do it
much better than we can. We just have to have open enough architecture, so
that maybe other people want to put our interface on their data structures.
And the reason they may want to do this is a couple of advantages we have
1) compile-time type safety, and
2) defining tables inside headers, which ensures that a program is in-sync
with the data structures.
Of cource we understand that there are limitations related to client/server
architecture, etc. We just have to find the appropriate niche for this
We would appreciate any further feedback.
Send and receive Hotmail on your mobile device: http://mobile.msn.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk