Re: Why is "group by" obligatory in SQL?
Date: Fri, 24 Jul 2009 11:49:48 -0700 (PDT)
Message-ID: <03c7d9f4-e6da-469d-bb79-b8d0f89f4143_at_h21g2000yqa.googlegroups.com>
Snipped
> >> What are the fundamental limitations of binary systems one should take into considerations when representing and operating relations.
> I'd say the very first one is that while we can 'see' or interpret a
> relation in the mind's eye, the machine representation, whatever form it
> takes, is by nature (and by definition where an implementation is
> concerned), idiosyncratic.
Yes but I was not refering only to informal limitations but also to
formal limitations such as how to represent a relation in a Ncoordinate
referential, be it a vectorial referential, a fractal based
referential or even a trigonometric referential.
Some of these limitations can be mathematically evaluated to quantify numerically the logical weight of operations. Such knowledge can for instance allow the quantification a direct image system relation representation vs a column store relation representation in function of degree and cardinality. This is the core of my work.
> >> How to effectively represent disjointness between two sets
> > ...
>
> Yes, given that representation is very much the basic decision for
> implementation. And what 'principles' (maybe axioms is a better word)
> could be adopted to interpret disjointness?
I would not call an effective algorhythmics for effectively
representing disjoint sets an axiom but I hear you. It is also very
difficult to exchange online on such complex subject, but let me try
express some premices I used so far.
--> A formal system that requires a logical ordering of elements to
represent disjointedness or some other property between 2 sets is a
direct image system.
--> Properties of disjointedness between 2 sets can only be expressed
if the 2 sets have common constraints that allow to define these sets.
--> Unordered representations must allow a numeric quantification of
logical IO's (a logical IO being the elementary constitution of a
single tuple). This quantification is a mathematical function that
involves relation cardinality and degree as well as the number of
permutations required to establish the property of disjointedness.
--> Disjointedness and other set level properties should have equal
logical IO's whether the members of the composing set ordered or not.
> How practical are the
> variious choices?
Unfortunately, practicality is a highly subjective concept. My guess
is that, o a fundamental level, the practical aspect depends on the
familiarity one has with mathematical models used to optimize the
establishment of the property of disjointedness. I have used 3 models
so far: vectorial representations, fractal representations and
trigonometrical representations. While I find fractals very practical
to make effective permutations for unordered sets, I prefer
trigonometric representations to express symetry between
disjointedness and its opposite intersections.
> Should an implementation assume disjoint sets that
> are eligible subsets of some relation are necessarily contrary assertions?
I am not totally sure what you mean by contrary assertion but if you
mean it as constraints defining the subset then the answer would be
no. One can only assume that the product of their intersection is the
empty set.
Received on Fri Jul 24 2009 - 20:49:48 CEST