Saturday, March 28, 2015

First Normal Form, Part 3: The Domain Imperative

(Continued from Parts 1 and 2)

As I explain in paper #6, a relation is defined on domains--it is a subset of a Cartesian products thereof. It follows that to be a faithful database representation of a relation, a R-table is also defined on domains--constrained data types, that is, types with constrained value ranges and applicable operator sets, which represent real world properties--from which columns--representing entity attributes--draw their values. Otherwise put, there are no R-tables without domains.

Saturday, March 21, 2015

Weekly Update

1. Quote of the Week
I'm not sure why you think integrity constraints are purely logical. Primary keys are physical constraints. They enforce that the primary key remains unique. Here's an example of SQL that creates a physical foreign key constraint.
ALTER TABLE FactInventoryCollections
   REFERENCES ViewCubeDimClient(ClientPK);
Physical constraints allow the database engine to return an error if an operation attempts to insert a row that violates any defined constraints.

2. To Laugh or Cry?
When One Data Model Just Won't Do: Polyglot Persistence

3. Online Debunkings

4. Interesting Elsewhere

David McGoveran has been working on a book tentatively titled LOGIC FOR SERIOUS DATABASE FOLKS intended to set some matters straight regarding the formal, set-theoretic and logic foundations of the RDM which have been misinterpreted. While he is not ready to publish yet, I asked and he agreed to post at his site a draft of a chapter on view updating which I consider a must read (together with the Introduction), particularly since it exposes the thinking behind the Principle of Orthogonal Design rejected by Date and Darwen.
David invites comments.

5. New Links

Added the following to the LINKS page:

6. And now for something completely different
Demis Hassabis, Founder of DeepMind Technologies and Artificial-Intelligence Wunderkind at Google, Wants Machines to Think Like Us | MIT Technology Review Demis Hassabis, Founder of DeepMind Technologies and ... The man behind a startup acquired by Google for $628 million plans to build a revolutionary new artificial intelligence ... At Harrah’s Casino on the shores of Lake Tahoe, DeepMind researchers showed off software that had learned to play classic Atari games including Space Invaders and Pong better than any human.
Can't get deeper than this!

The PostWest
In my native tongue there is a saying: "If somebody pees on you, don't pretend it's raining". Hard to tell the super- from the third rate power, ain't it?

The oldest hatred
Upside down and backwards: It's not the  US that wreaks havoc in the ME via Iraq, Afghanistan, Egypt, Syria, Lybia and IRAN, no, it's--you guessed--Israel that causes chaos! Anti-semitism is a blinding sickness.
Middle East Reality Check

This week's video
Self-Organization: Turing & Darwin

This week's book
The Great Reformer: Francis and the Making of a Radical Pope

Saturday, March 14, 2015

First Normal Form, Part 2: Some History

(Continued from Part 1)

Consider now the email's author reply to the question in the exchange to which he referred me:
Re "atomic": In Codd's original 1969 and 1970 papers he explained that "atomic" meant not relation-valued (ie not table-valued):
So far, we have discussed examples of relations which are defined on simple domains--domains whose elements are atomic (nondecomposable) values. Nonatomic values can be discussed within the relational framework. Thus, some domains may have relations as elements.
He used "simple", "atomic" and "nondecomposable" as non-relational informal expository notions. He understood that a relation has rows of which each column has an associated name and value, and whatever ("single") value is put in is what comes out. The only structural property that matters relationally is being a relation. It is also just a value, but you can query it. Then he used "nonsimple" etc. meaning relation-valued.

Saturday, March 7, 2015

Weekly Update (UPDATED)

1. Quote of the Week
I have been in the data side of IT for quite some time now and have seen the evolution of how data is ingested, manipulated and regurgitated to the end users in hope of telling our consumers "how much of something did something". The main issue seems to be complexity of the data models and the fact we don't have a model that can expand with the data without adding tons of new schema. The solution.

2. To Laugh or Cry?

3. Online Debunkings

4. Interesting Elsewhere

5. And now for something completely different

The PostWest
  1. Zarif: The Netanyahu regime 'should be annihilated'
  2. Iran’s Shiite Drive Scoops up Yemen en Route from Syria & Iraq
  3. Allies may see Iran deal as sign of superpower in retreat 
  4. Turkey joins Saudi-Pakistani Sunni Nuclear Partnership versus Iran’s Shiite Nuke 
  5. US shifted from preventing nuclear Iran to containment 
  6. Iran brings Europe within range with new cruise missile

What exactly was factually inaccurate in Netanyahu speech, again?

And don't believe the mainstream press:

The oldest hatred in the New World

Arab-Israeli Conflict reality check

This week's video

This week's book

Thursday, March 5, 2015

Domains, R-tables, and SQL

March blog post @All Analytics:

To ensure sensible results from and correct interpretations of analysis of data from SQL tables or extracts thereof, analysts must know the tables’ interpretation -- the business rules underlying them -- which is rarely documented.

They should be represented in the database by integrity constraints -- not perfect substitutes, because they are very loose approximations to the rules -- but if they are enforced in the database by the DBMS they are usually recorded either in the definition statements that created the tables and constraints, or the database catalog.

Read it all

Sunday, March 1, 2015

First Normal Form, Part 1: Atomicity

Shortly after posting First Normal Form Is Not Structural Regression I received the following email, taking me to task for stating "[Codd's] initial concept was not "single-valuedness", but "atomicity", which was problematic, as it lacks a precise definition."
Contrary to this received wisdom, Codd in 1969/70 (and RM V/2 20 years later) gave a precise relational definition of "atomic" aka "simple" aka "nondecomposable" (later aka non-"compound" aka non-"structured"): not relation-valued. And he gave a precise definition of "normalized": free of relation-valued-domains.
I was referred to an exchange in which the argument is further amplified.

Wednesday, February 25, 2015

SQLSaturday Presentation

March 28, 11:15,  Mountain View


You're a SQL Server ace: your ability to squeeze everything from SQL and your performance tuning skills are unparalleled, but do you know what your tables really mean and, therefore, what queries make sense and whether results are correct and their interpretations sensible? This is a critical part of data fundamentals, the grasp of which is poor. It is a subject usually neither much covered in education, nor part of job requirements and industry dialogue, yet can defeat the entire purpose of your DBMS expertise. This presentation covers
  • Meaning, business rules and table interpretations;
  • Types of business rule; 
  • Meaning and database truth; 
  • Business rules, integrity constraints and database consistency; 
  • DBMS and user reponsibilities.
Session Level: Intermediate

Event full details

Contact: Mark Ginnebaugh