Re: [Monetdb-developers] Object persistence using MonetDB

darabi@web.de wrote:

Thank you for the information.

So, does that mean that you generally do not recommend the use of MonetDB beyond the realms of

...

datawarehouses, information retrieval, and data mining ?

From the result of the web application contest in the ct magazine, I had the impression that 'normal' applications could also benefit from the architecture of MonetDB:

http://sourceforge.net/forum/forum. Indeed, with some careful application programming you can get a lot of

No not at all, but you should not expect wonders. traction.

php?forum_id=584739

What I would like to do is to create a persistence engine for a Prolog like language which has a type system. Which means that data is to be stored in tuples.

Prolog resolution engine is tuple based. Depending on the interaction method, you may well benefit from its binary table approach, but this should be done on top of the Prolog execution model. Backtracking over individual elements stored in MonetDB is not a good idea. Think about how to exploit the bulk processing power in you algorithms.

To experiment with MAL, I would have to use the MapiClient, right? I think I will try to play with it.

Good. Please keep us posted on progress.

Thank you

Kambiz

----- Original Message ----- From: "Martin Kersten" To: darabi@web.de, monetdb-developers-bounces@lists.sourceforge.net Sent: Mittwoch, 10. Oktober 2007 11.01 Uhr (GMT+0100) Europe/Berlin Subject: Re: [Monetdb-developers] Object persistence using MonetDB

darabi@web.de wrote:

...

Hello,

I'm a developer located in Germany with a current interest in object persistence for a dynamically typed language.

Hello, thank you for your interest in the MonetDB software.

...

I was experimenting with something along the lines of the decomposed storage model when I found out that it was (of course) decades old and found the papers about it and MonetDB.

Many thanks for your efforts and for this great piece of software.

I was reading the 1995 CWI report 'High Performance support for OO traversals in monet' and was wondering whether the ideas in it are still present in today's MonetDB. And if not, how much effort would it take to implement OO persistence using MAL. I would also appreciate any pointer to papers, code samples etc. in this direction.

Between 1995 and 2003 MonetDB was used in the company Data Distilleries (now part of SPSS) where the interaction between application and DBMS was based on persistent objects. However, the thrust of the development is not on individual object access, but bulk processing as it emerges from datawarehouses, information retrieval, and data mining.

...

If I understand correctly, in the beginning, MonetDB was rather heading towards a OODBMS. Was there any reason why you didn't pursue that goal? Is there any road block?

There are (have been) attempts to hook up Hibernate and MonetDB. I don't know their status. If ultra-high performance is your target, I question if a DBMS is the right way to go. However, I also think that most o-o applications might better benefit from relational view on data. regards, Martin

...

Thank you very much for your help.

Kambiz Darabi