10 Mar
2008
10 Mar
'08
5:02 p.m.
Hi Hani,
I would use Mondrian.DistributionMap that you have
recently
implemented, but the pb that I had is:
-each distributed entity has at less two properties (let us imagine:
'first property'->'fill color darkness' and 'second
property'->'border
color darkness').
-in some models may be we will have more than two properties to map:
we can have 4, 5, etc. of properties but in the same time any given
entity could not be attached to more than two properties (in other
words we need to mapping different properties to different colors).
The Distribution Map is an effective technique for the exact problem
of comparing two partitions with each other:
- one partition is represented in the containment (rectangles contain
squares)
- one partition is represented with colors
If you add more partitions, I am not sure that just taking the
distribution map metaphor as it is will work. In particular, I am not
sure that it would work if you start to play with the border, because
the picture will be too crowded to be useful.
-could we have the possibility to use different shapes
for the
distributed entities (some entities are boxes while some others are
ovals).
Maybe this could work, but then it will be a bit difficult to come up
with enough distinctive shapes.
Now could you, please, tell me how to improve the
actual
implementation of distribution map? may be we should implement a new
distribution map, what you think about the time cost?
The Distribution Map is not much code and does not require much
effort. I would suggest to play with it in your code and then if
something better comes out of it we push it in Mondrian. Is that Ok?
Cheers,
Doru
--
www.tudorgirba.com
www.tudorgirba.com/blog
"From an abstract enough point of view, any two things are similar."