On 3 mars 2010, at 15:38, Cyrille Delaunay wrote:

Hi,

I'm looking to do this customizable DistributionMap and I have several questions:

-> When you say 'customizable', I understand that what should be customizable are: -> kind of properties we want to display -> Color associated to each property Am i right? I guess the kind of the parts (Packages , classes ...) and elements (Classes , methods ) to display, will depend on the moose-group we first clicked-on (to launch to visualization).

-> What kind of property can be display in a DistributionMap? In the distributionMap paper, I saw that , for example, properties could be 'the owner of a file', or also the result of a clustering algorithm like 'linguistic concepts'. What kind of property should be propose by the wizard?

-> Is it better to have a wizard-based customizable map or a full-dynamically-customizable visualization like the one of SystemComplexity ?

Indeed, it's easier to do a wizard in this case, simply because we dont know which property we want to visualize (contrary to metrics in SystemComplexity). For colors, it's a bit different: I do remember something about the order of colors and I think a default set is good enough.

Take a look at MODistributionMap, it uses blocks to actually customize parts and properties based upon the group. You can also take a look at DistributionMapExample, a simpler but working distribution map (no serialization).

MODistributionMap>>initializeWith: aGroup property: anotherBlock partitioning: aBlock self createElementsFor: aGroup. self createPropertiesWith: anotherBlock. self createPartsFor: aBlock

So I guess that all you is some text fields to enter a block or a symbol to extract parts and properties

-- Simon