Needle's "Blank" Data Format

Written by Chuck McCallum Tuesday, 31 August 2010 17:42

In Needle, the underlying data is a graph: "nodes" of data are linked to other data nodes. If you're used to thinking of your data as tables, either single spreadsheets or relational tables in a database, this can be an adjustment. After you're used to it, though, a graph representation will often feel more natural. In cases where it isn't, a table can still be straightforwardly represented as a graph.

There are some cases where the graph representation isn't immediately obvious. Take this example:

Date	Time	Home	Away
Sept. 4	7:30pm	D.C. United	Columbus
Sept. 4	8:30pm	FC Dallas	Toronto FC
Sept. 5	8:30pm	Houston	San Jose

This is a portion of the Major League Soccer schedule. It's about games, but there isn't anything by itself that can stand for a distinct game: Games don't have names. Instead, in Needle, we would create a "Game" type with a "blank" data format. "Blank" means it has no intrinsic value of its own, and is instead defined by its links to a "Date/Time", "Home", and "Away". Were it not for double headers, the time wouldn't be essential to the definition of a game.

Two new sample domains are good examples of the versatility of "blank" data formats: U.S. nuclear reactors and the CDC's vaccine recomendations for travellers. If you look at the domain model for the CDC domain, you'll see that "Vaccine Recommendation" has a "blank" data format and is related to a single "Vaccine or Disease", a single "Explanation", and any number of countries:

• Vaccine Recommendation:
  • Vaccine or Disease
  • Explanation
  • Country (multiple)

The recommendation for a vaccine may vary, depending on the destination, but these distinct recommendations do not have distinct names. Instead, a distinct vaccination-explanation pair is what constitutes a "Vaccine Recommendation". Since they do not have an intrinsic name, and since the "Explanations" can be long, when listing "Vaccine Recommendations", we just show the related Vaccine.

The U.S. Nuclear Reactors domain is similar: When you look at its domain model, note that "Reactor" has a "blank" data format, and is linked to a single "Plant" and a single "Unit Number".

• Reactor:
  • Plant
  • Unit Number

On the NRC site, much of the data is oriented around individual reactors, but the general public is more likely to be interested in the characteristics of the plant as a whole. Needle's flexible data model makes it easy to pull from a source centered on one type (here, the reactors), and display the data as if another type were the center (here, plants). At first, it might seem that what's been tagged as "Unit Number" could have been the "Reactor": in many cases that would work, but in others no unit number is mentioned, because a facility consists of only a single reactor. When present, the "Unit Number" distinguishes a "Reactor", but at a single reactor facility, it is optional.

< Prev		Next >