The multiple links idea is a bit unusual in the context of what I've
read about for graphs but all three methods I've mentioned could be
adapted. However, given the additional constraint of per-link
information, perhaps the adjacency matrix is not the way to go but you
could use this and just have a link information list of boxed items
with multiple items for multiple links.
So, assuming a graph like this (where the "V<>" characters are meant
to be arrowheads):
0->1->2
|\ |
V \ V
3 \__>4
| /
V /
5 <__|
where the link from 4 to 5 is really 3 links, and the 0 to 1 is 2
links, a vertex array might look like this:
VA=. ,&.>(1 1 3 4);2;4;5;(5 5 5);<i. 0
LIva=:
,&.>('rel1';'rel2';'rel0';'rel1');(<'rel0');(<'rel3');(<'rel1');('rel0';'rel2';'rel3');<''
Here the "from" link is implied by the index into VA; LIva is the link
information grouped by "from" node; the ",&.>" ensures we have all
vectors.
The LIva array could also be used with the existing boolean adjacency
matrix: the number of links for node "n" would be implied by #>n{LIva
.
A node;edge representation would be like this:
NE=: (0 1 2 3 4 5);<0 1, 0 1,0 3, 0 4,1 2, 2 4, 3 5, 4 5, 4 5,: 4 5
LIne=: 'rel1';'rel2';'rel0';'rel1';'rel0';'rel3';'rel1';'rel0';'rel2';'rel3'
So, this shows how the links and relationships line up:
}.NE,<>LIne
+---+----+
|0 1|rel1|
|0 1|rel2|
|0 4|rel1|
|1 2|rel0|
|2 4|rel3|
|3 5|rel1|
|4 5|rel0|
|4 5|rel2|
|4 5|rel3|
+---+----+
Notice that the VA representation explicitly includes nodes with no
children whereas the NE one only implies them. Also,
LIne -: ;LIva .
On Wed, Mar 21, 2012 at 3:37 PM, Alexander Mikhailov <avm...@yahoo.com> wrote:
> -----
>
> Date: Wed, 21 Mar 2012 09:25:19 -0400
> From: Raul Miller <rauldmil...@gmail.com>
>
> The "natural" J data structure here would be two items:
>
> 1. A list representing data at each node.
> 2. A connection matrix.
>
> [And I saw some other messages in this thread advocating this kind of
> approach -- there's good reasons for that.]
>
> -----
>
> How to store link-associated data then? And how to specify that there are 3
> links from node 4 to node 5?
...
--
Devon McCormick, CFA
^me^ at acm.
org is my
preferred e-mail
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
