Feedback

[guilhelm savin <guilhelm.savin AT graphstream-project.org>]

It seems a bit distribution problem happens, so I transfer this mail:

---------- Message transféré ----------
From: Michael Szell <michael.szell AT meduniwien.ac.at>
To: users AT graphstream-project.org
Date: Fri, 17 Sep 2010 13:36:11 +0100
Subject: feedback

Hello,

this is a list of feedback on issues/feature suggestions for making
life easier for the gs user which I have encountered after my first
day of use. From what I have seen at the ECCS'10, gs has the potential
for becoming a standard tool for the visualization of evolving
networks, and the developers have done excellent work so far. Note: My
background is in maths/physics, my work is on social network analysis
(so I don't know anything about biological or other networks), and I
have no experience with Java.

1) Real time: Many datasets (e.g. mobile phone calls) come with
timestamps, and since human behavior is kind of "bursty" (very
heterogeneous in time), a feature is needed to incorporate this
real-time information, at least for the visualization (the "steps" of
gs are too poor for this). Here an example of such data:
SourceNodeID DestinationNodeID Timestamp(UNIX-time)
9993659 82423521 1182187454
9993659 82423521 1182193682
9993659 82423521 1182193693
51494778 14040501 1182195482
51494778 14040501 1182195500
51494778 14040501 1182195506
51494778 76170314 1182195560
....
A simple solution could be to give st an optional attribute "time",
then one could translate+include timestamps into the dgs format.

2) More tutorials: For example, a step by step setup of eclipse+gs
with a "hello world" example (as in the real-life ECCS tutorial), how
to visualize multi-edges, or how to export an animation into a movie
file.

3) Relation types: In social networks, there are persistent relations
(e.g. friendship), which remain over some time and can eventually be
deleted, and there are one-time events such as writing an email. In
network studies these one-time events are usually accumulated over a
time window of some days, since e.g. an email you wrote to someone a
year ago is not relevant anymore. It would be great if there was some
function which adds a link, immediately deletes it, but in the
visualization it is shown as some kind of disappearing arrow (see
http://www.youtube.com/user/complexsystemsvienna#p/a/u/0/NC-Gtu2yaCU
for an example). It could be realized by a new datatype, such as
"OneTimeEdge".

4) A "sloppy" option (boolean parameter in addEdge() and
nextEvents()?) for handling of edge additions: Whenever one tries to
add an edge between non-existing nodes, an exception is thrown
("cannot make edge from 'A' to 'B' since node 'A' is not part of this
graph"). As a user I know that these nodes have been newly added with
the edge, and I don't want to bother adding them by hand. Removal of
nodes is already "sloppy": When I remove a node, edges connecting to
it are removed too - I don't need to remove the edges first (or do
I?).

5) Pajek format: Most people in my field use Pajek for visualizing
static networks. Support of the Pajek network format would maybe reach
more people (personally I think the format is ugly, but many people
use it).


I would love to use gs to create some videos with my data, but at the
current development stage it seems I would have to spend too much time
trying to find out how things work in gs and java, and writing
hacks/workarounds. But maybe in the near future! :)

Cheers,
Michael

--
Section for Science of Complex Systems
Medical University of Vienna
http://www.complex-systems.meduniwien.ac.at/people/mszell