Hi,
On 20 Jun 2011, at 15:11, Nicolas Anquetil wrote:
Guys,
For what it is worth, may I suggest that this discussion is going nowhere? If each one keeps repeating the same argument this is going to be a very long argument with nasty side effects (not on the software).
I do not mean to annoy anyone. I actually thought we were actually making progress (no joke). I think we are all trying to explain what we think, but I think there is a mismatch in the points of view, and somehow it just does not seem to come out right :)
I am interested in getting it right, because it can produce a nice documentation, or it might uncover some other solutions.
Now, may be the problem is that the "input to Mondrian are not nodes & edges"? Because it is based on a graph model and we do use it to represent graphs, so naturally, one expects to use it that way ... The next questions are: What does it use as input? And why is it not just plain edges&nodes?
Because you do not always have a natural set of edges in your entity model. For example, in Smalltalk, there is no object that represents inheritance, and yet you would want to represent inheritance as an edge visually. Visually :). So, to do that, you will take the class as the model of the edge and draw the edge from that class to the #superclass.
Now, if you do have an object that represents a relationship, then by all means use it as the model behind the edge. For example, for the same example of inheritance, in FAMIX we would use FAMIXInheritance.
In other cases, you have only relations without natural nodes. For example, if you have a collection of mails that have senders and receivers as strings, the mail is a relation, but the nodes need to be constructed.
The idea of Mondrian is to not limit the possibility of visualizing to only graph models. That is why you have blocks as transformations.
I think, that the answer "The input are only entities which can be anything" is not satisfying because users will naturally be thinking in terms of what they want to see in the end: nodes&edges
How is it now?
nicolas
PS: By the way, the buggy comment is mine, I proposed it when first trying to understand how things worked. I, also, did not find it very intuitive that edges: does not simply take a collection of edges as parameter.
No problem :)
Doru
I am repeating for the n-th time: The input to Mondrian are not nodes & edges. The input are only entities which can be anything. The engine simply provides a means to transform your entities into nodes and edges.
Take this point of view, and you will see that it becomes consistent :)
You keep on repeating that you are passing nodes to the edges:, but you are not. You are always passing some entities from which the edges are to be constructed. That is all. It just happens that you choose to use the same entity as a model for both an node and an edge. It is clear that if there is no easy mapping between the entity model and the graph model, the script can become cumbersome.
Let's take your example: view node: 'node' forIt: [ view interaction whenEnteringUpdateNode: [ :aValue | view edges: {aValue} from: #yourself toAll: ((1 to: 9) copyWithout: aValue ) ].
view shape rectangle size: 30. view nodes: (1 to: 9). view gridLayout gapSize: 10.
First, this example is not instructive because it takes several shortcuts, but let's just focus on aValue.
In your script, aValue is the value behind the node, not the node. The node is another object. Then you use it also as the model for the edge because you do not have anything else. So, you will have the same aValue object used both for creating the node and for creating the edge.
So, a method like buildEdgesFromVertices: nodes fromAccessor: aValuable toAccessor: aValuable is wrong in the context of Mondrian, because it uses the wrong vocabulary.
How is that? :)
Cheers, Doru
On 20 Jun 2011, at 14:15, Stéphane Ducasse wrote:
On Jun 20, 2011, at 1:32 PM, Tudor Girba wrote:
Hi Stef,
On 20 Jun 2011, at 13:18, Stéphane Ducasse wrote:
I think that this is more than just having correct/good comments....
is the parameter passed to edges: consistent? I do not think so
It is consistent. All the time the first argument sent to edges:from:to and nodes: is meant to hold a collection of entities that will be the model behind the graph element. In the case of nodes, there is no fancy transformation. In the case of edge, there are two transformations. That is all.
If you find a case in which this is not true, then we need to fix that.
at least from what I understood
edges: from: to: and aView edges: someFamixInvocations from: #sender toAll: #candidates
show different cases.
Now if nodes: except a collection of entities playing the node role why edges is not a collection of antities playing edges role?
It is.
No it is not You have to send messages to nodes to get edges so the collection is not an edge collection
Corollary why in some cases this is the case and in other not?
Just the comments are misleading. It is always the same. Only when you have something marked differently, you expect a different input, like edgesFromAssociations: which expect to have a collection of associations. But those are just convenience.
I really lost then. Because in
view node: 'node' forIt: [ view interaction whenEnteringUpdateNode: [ :aValue | view edges: {aValue} from: #yourself toAll: ((1 to: 9) copyWithout: aValue ) ].
view shape rectangle size: 30. view nodes: (1 to: 9). view gridLayout gapSize: 10.
aValue is a node not an Edge?
I always have the problem with mondrian that what I say is wrong. It is a wonderful feeling.
Now I see what the API says
is Oh nodes: nodes you specify nodes with nodes nice
but you specify edges with nodes...... not nice
edges: nodesAnd for me this is confusing. Because why do I have to remember that edges: is not about edges but about nodes that will generate information about edge. What you call declarative for me is really confusing.
Now changing aCollection to aCollectionOfEntities only work is you explain systematically in the comments that theses are entities in fact the entities that are not first class relationship and nodes that can generate information to represent the edges. Else without context collectionOfEntities look like collection of objects.
Again, the comment needs to be fixed.
Does this clarify the situation?
Yes but I do not like it. Why because
edges: from: to: is connoted
Since math exist edges: are edges and not vertices from: is a vertex and to: another vertex
So if I need to build visualization in Mondrian I will build a layer on top a layer for idiots like me.
buildEdgesFromVertices: nodes fromAccessor: aValuable toAccessor: aValuable
Now I would really like to have a non declarative way of expressing that. In fact I do not even understand why a declarative way is supposed to be better than an imperative one. Especially because most of the time the order is important in Mondrian.
Declarations does not talk to me and the the stack semantics of Mondrian is difficult to understand and debug. And do not dream there will be no debugger able to debug Mondrian scripts in the close future. So we are left with method names and arguments and guessing right.
What frustrate me with mondrian is that I'm programming in lisp without any debugger, but worse not printf..... So I start fast and boum I hit the wall and got trapped in DSL swamps.
Stef _______________________________________________ Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"There are no old things, there are only old ways of looking at them."
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"Obvious things are difficult to teach."