On Mon, Jul 24, 2006 at 09:13:28AM -0700, hanno wrote: > Subject: Re: [sqlite] selecting rows of the view via its position
> My real problem is, that I do all that stuff for my diploma's thesis in > computerscience. I'm trying to show the abilities of relational datamodels > in a deductive context. I'm trying to implement a rule base in sql that > shall provide correct and complete analysis of chess situations as an > example for deductive database driven monitoring. Well, I'm no expert on this, but keep in mind that "Standard SQL" != "relational model". They are related, but not necessarily the same. (In many of his writings, C.J. Date complains in detail that SQL is not properly relational.) What you're trying to do sounds related to the (older) research work on "deductive databases", e.g., Aditi. All the various forward and (particularly) backward chaining systems are probably at least peripherally relevent. E.g., the Prolog, Mercury, and Oz languages; the Oz/Mozart and CLIPS toolkits; and various others too. http://www.cs.mu.oz.au/research/aditi/ http://www.cs.mu.oz.au/research/mercury/ http://www.mozart-oz.org/ http://www.ghg.net/clips/CLIPS.html http://openacs.org/forums/message-view?message_id=44805 Also, (as far as my limited understanding of they underlying theory goes) the relational model is fundamentally un-ordered. I'm told there is some academic work out there on other, more powerful models which natively understand order. I think it would make more sense to call that a "vector relation model", but apparently what it's actually called is a "set model"! I have not read it, but this book seems to cover that subject: "The Set Model for Database and Information Systems" by Mikhail M. Gilula http://www.amazon.com/gp/product/0201593793/102-7773671-0297757 Some of the gurus of the vector-oriented languages (APL, K, J, A+, etc.), particularly the K folks, have implemented relational-like database systems. I strongly suspect they grok this un-ordered vs. ordered distinction, and are taking advantage of their vector languages' native understanding of and optimization for ordered data. I don't know whether or not that's related to Gilula's "set model" above, though. Ordered data is perhaps also related to column-store (rather than the more usual row-store, like SQLite, Oracle, etc.) databases. MonetDB is one such (and OpenSource) column-store database. -- Andrew Piskorski <[EMAIL PROTECTED]> http://www.piskorski.com/
