Data Laboratory

[admin]

This is the thread for asking more details about the Data Laboratory proposal.

[sumitd1234]

I have a question regarding "Data Laboratory" Project. For things like search and autocompletion etc, for a editor I used a package RSyntaxTextArea, can such a tool be used in this project also or these things will have to be coded separately.

[mbastian]

I have a question regarding "Data Laboratory" Project. For things like search and autocompletion etc, for a editor I used a package RSyntaxTextArea, can such a tool be used in this project also or these things will have to be coded separately.

I don't get how RSyntaxTextArea could be used in this project. The proposal says

Search on one or all columns. Regex based search. Autocomplete if possible.

The autocomplete would be done on the search field, thus nothing like an editor.

[sumitd1234]

Hello,

As I discussed with mbastian on IRC, values of the attributes in a column can be accessed only through going through all the rows. So should I aim to maintain a new data structure for attributes in columns or go for searching though the rows(nodes) to provide autocompletion ?

Thank you,
Sumit

[mbastian]

Yes that is correct, attribute values are stored in the AttributeRow object, which is accessible only from nodes or edges. Therefore rows are not maintained in the attribute model. It would cause a conflict with GraphAPI, that already maintains the list of nodes.
For autocomplete and search requirement the attribute modules would have to be extended with a reverse index structure. This structure has to be cleanely defined and scalable.

[sumitd1234]

Hello,

I have a two doubts regarding undirected graphs. When I opened a dataset of undirected graph http://gephi.org/datasets/power.gml.zip , current data laboratory module still shows source and target for edges. I mean shouldn't this be like node1 or node2 as is defined in GraphAPI for UndirectedGraph.java or am I missing something?
And so symmetrizing as I see now is like one of two possible options
a) creating and loading a new dataset( where the source and target nodes are converted to node1 and node2 and opposite edges are merged to one( we dont have parallel edges, right?)
OR
b) using the same dataset just temporarily moving the edges in edge iterator of directed edges to edge iterator of undirected edges and flagging the graph as undirected. Am I understanding the problem correctly?

Thanks,
Sumit

[mbastian]

I have a two doubts regarding undirected graphs. When I opened a dataset of undirected graph http://gephi.org/datasets/power.gml.zip , current data laboratory module still shows source and target for edges. I mean shouldn't this be like node1 or node2 as is defined in GraphAPI for UndirectedGraph.java or am I missing something?

Yes, the columb label should be changed.

Undirected edges are fullt supported by the GraphAPI implementation. Iterators just ignore mutual edges when iterate over edges. The graph is flagged undirected if only undirected edges have been added to the graph. It's directed by default and mixed if both directed and undirected edges are appened. Regardless edges type it's always possible to get directed or undirected graph from the GraphAPI.

[sumitd1234]

Hello,

I was going through the sparklines and their use in things like dynamic network analysis. Dynamic attributes http://gexf.net/format/dynamics.html , I guess are those which have one or more slices of start-end time and different values per time slice. So what exactly am I supposed to draw in the sparklines, the values of the dynamic attributes of the nodes as a function of time? Similarly for edges also, or edges have to be dealt separately?

Thanks in advance

[mbastian]

So what exactly am I supposed to draw in the sparklines, the values of the dynamic attributes of the nodes as a function of time? Similarly for edges also, or edges have to be dealt separately?

Both nodes and edges can have dynamic attributes. Sparklines would just display values as a function of time indeed. Note that these values can be the data, imported or have been computed, for instance the in-degree.

[sumitd1234]

Hello,

For the scatter plot feature in data laboratory, I found JFreeChart which is an open source package for creating plots and charts. Will it be a feasible idea to include it while developing data laboratory as integrating this package may be useful in long run as it is really good at creating various kind of plots http://java-source.net/open-source/char ... -reporting . If mentors dont think it to be a good idea, then I think I will have to propose to do it the swing's way but still isnt it a good place to learn developing plots?

Thanks