On 24/03/13 22:36, DoanVietTrungAtGmail wrote:
I read your help (thanks, Peter and Alan) and thought that relational
database technology didn't naturally map to my problem domain. Here, at
each simulation timestep each entity has a run-time variable number of
pointers to other entities. Representing each entity as a table seems
unwieldy to me.
It really depends on your problem.
If you have a great many number of entity types then relational may not
be the best solution. If you have a lot of entities of a few types then
relational is a good match - thats what relational databases do best.
So the question is how many data types you need to deal with and the
dependency graphs between the types. If there are many types all cross
referencing each other then a non RDBMS may be a better fit. (although
you can get round it in an RDBMS by modelling the type relationships as
tables and then simply storing references to the type relations, but
that's a bit of a forced solution i agree)
One possible set of solutions is the family of "NoSQL databases
There are a huge number of these ranging from single "big table" models
to network and graph databases. There are also a myriad of recent "big
data" stores too, with their own query languages, for example I've
played with Hadoop.
Which one suits your problem depends on what you need to store and what
you want to retrieve after you've stored it!
--
Alan G
Author of the Learn to Program web site
http://www.alan-g.me.uk/
_______________________________________________
Tutor maillist - Tutor@python.org
To unsubscribe or change subscription options:
http://mail.python.org/mailman/listinfo/tutor
