Re: [xdi] On current literal pattern...

If you want to retrieve only literals, isn't that what Mike's $get$literal proposal is meant to do?

To be honest I like neither $!(+tel) nor +tel$!
Sorry, maybe I missed or forgot the discussion on why something like that was introduced. Why can't it just be +tel

Markus

On Thu, May 24, 2012 at 3:09 PM, Barnhill, William [USA] <> wrote:

Assume a graph contains result of adding the properties from the simple example in the most recent XDI Graph Patterns doc:

1 () () (=bob)

2 () () =bob

3 () () =alice

4 =bob () $!($uri)

5 =alice () $!(+tel)

6 =alice$!(+tel) () $!($t)

7 =alice +friend =bob

8 =bob$!($uri) ! (data:, http://xdi.example/com/=!0111.7af3.65d5.8cb7/)

9 =alice$!(+tel) ! (data:, +1-206-555-1212)

10 =alice$!(+tel)$!($t) ! (data:,2010-10-10T11:12:13Z)

Assume I want to traverse the graph to get all literal properties under =alice at any depth, a situation that could occur due to serialization, or processing for an XDI IDE.

Assume I want to do this solely by using graph operations, due to whatever algorithm I have.  In other words I can walk the graph and test equality of a node or edge's label against a specific label, but I can't use a REGEX or parse the label.  I think this is a reasonable assumption, and I think if we can't make it then we don't strictly have a *graph* model.

Can I still get all the literals by walking the tree if I follow these assumptions?

My concern is with the use of things like $!($uri) and $!(+tel).

As a result I propose the following pattern instead of those statements:

1 () () (=bob)

2 () () =bob

3 () () =alice

4 =bob () $uri

5 =bob$uri () $!

6 =alice () +tel

7 =alice+tel () $t

8 =alice+tel () $!

9 =alice+tel$t () $!

10 =alice +friend =bob

11 =bob$uri$! ! (data:, http://xdi.example/com/=!0111.7af3.65d5.8cb7/)

12 =alice+tel$! ! (data:, +1-206-555-1212)

13 =alice+tel$t$! ! (data:,2010-10-10T11:12:13Z)

Now I can find all literals under =alice by finding any walk from =alice to $! within the current graph.  This also seems easier to parse, both for a human reading it and a software agent.

Furthermore, I think we shouldn't use literals to represent a URI.  A URI can be represented as an XRI just by wrapping it in cross ref. So I'd also suggest changing the following lines:

4 =bob () $uri

5 =bob$uri () $!

11 =bob$uri$! ! (data:, http://xdi.example/com/=!0111.7af3.65d5.8cb7/)

  to ...

=bob $uri (http://xdi.example.com/=!0111.7af3.65d5.8cb7/)

The complete result becomes

1 () () (=bob)

2 () () =bob

3 () () =alice

4 =bob $uri (http://xdi.example.com/=!0111.7af3.65d5.8cb7)

5 =alice () +tel

6 =alice+tel () $t

7 =alice+tel () $!

8 =alice+tel$t () $!

9 =alice +friend =bob

10 =alice+tel$! ! (data:, +1-206-555-1212)

11 =alice+tel$t$! ! (data:,2010-10-10T11:12:13Z)

Note that I also changed the URI in 4 to have an authority ending in .com (was example/com/), and removed the trailing '/'.

---------------------------------------------------------------------

To unsubscribe, e-mail: 

For additional commands, e-mail: