datalang-spec — DL specification working group

[Tom B. <bra...@ap...>]

On Tuesday, March 19, 2002, at 03:31  PM, Mathieu Bouchard wrote:
> Vector and Array are different classes; creating a Vector of elements of
> identical type will encode as something that will be parsed as an
> Array; a similar thing happens between Map and Array.
>
> I'm not sure why there is a distinction between Vector and Structure.

[Took it off XML-Dev... doesn't belong there.  We should move this to 
the DL spec mailing list on SourceForge]

A Vector will only report values and other structures, but not 
whitespace.  In this way, it behaves more like a true programming 
language data structure.  Also because it can more easily be determined 
whether or not the structure has passed through a validator.

> Will there be user-defined constructed value types? Will there be
> constructed value types taking several values?

Not sure yet.  I'm at version 0.1 and still in the planning and design 
stages of the project, so this question probably deserves more in-depth 
discussion.  I'm prone to say that the simpler, the better, but there 
has to be a balance between simplicity and utility.

> Where else is that particular date format used? Why was it chosen?

It's ISO 8601

> Why are 'date' and 'uri' considered to be universal enough to be 
> included
> in the base set? How do i convert a org.notdotnet.dl.objects.Vector 
> into a
> java.util.Vector? into an Object[]?

Again, I'm at such an early stage that these things are still up in the 
air, but I wanted to prove a use case where a value could only be 
constructed in a more generic fashion.

For the moment, Vectors, Maps, and Arrays won't even work because there 
is no supporting validator that will generate them.

> Why are constructed value types different from maps? why is
> there no distinction between a hash key and a map's name ?

The reason for the difference is that a constructed value yields a 
single value (object, token, whatever), and a Map yields a structure 
with a single nested value.  The term Map will likely change, as it's 
somewhat confusing.  Really what is being described is a struct-value 
pair.

> Could the DL library be closer to a serializer in the sense that it 
> would
> operate on hierarchies of objects created by programs that don't inherit
> from the DL library ?

This is a later-stage goal, but will likely be implemented as an adapter 
rather than as part of the system itself.

> And to speak of name conflicts at another level: Map is a word often 
> used
> to mean dictionary or function; in some languages, Vector is the name 
> for
> single-dimensional Arrays, and Arrays can have multiple dimensions; in
> some others (Java) Array is an optimisation on Vector, and appears to be
> type-restricted, except type-restriction can be almost arbitrarily loose
> (Object[]).

So any recommendations?  These names do overlap with the Java usage of 
the terms, partially to ease people into the concepts, but they're not 
set in stone.

Let's move this to the DL spec or devel list.

--
Tom Bradford - http://www.tbradford.org
Architect - XQRL (XQuery Engine) - http://www.xqrl.com
Apache Xindice (XML Database) - http://xml.apache.org/xindice