Re: Constraints and Functional Dependencies
Date: 24 Feb 2007 21:52:19 -0800
Message-ID: <1172382739.510914.162210_at_z35g2000cwz.googlegroups.com>
On Feb 24, 4:12 pm, paul c <toledobythe..._at_oohay.ac> wrote:
>
> Folklore has it that it's a big mistake to use RA, at least a stepwise
> version of it, for writing constraints because an engine will have
> little choice but to apply it verbatim.
Interesting. I had come to the same conclusion; good to hear it confirmed, if only by folklore.
> Whereas an engine will likely
> find it very convenient to convert whatever declarative constraint
> language to something resembling RA. However the same query might
> result in different generated RA on two different days.
Yes. Sadly, there is no way to canonicalize a query. I thought for a while I could come up with one for the Relational Lattice, but my approach didn't work. My expectation is that it's undecidable. It's known to be undecidable for conventional RA, and the RA is translatable into the RL, so we would certainly expect it to be undecidable there as well.
Marshall Received on Sun Feb 25 2007 - 06:52:19 CET