[FIXED] The dangers of using Ranking

[rotten]

I have observed that when one applies a scale, in particular on edges, the relevant Data Table column gets updated - and the original values appear to be lost.

The problem is if you decide the scale is not a good one - you want to try a different Spline, or a different Range, you have to reload your data, because you've already trashed your original column.

It took me a long time to figure out what was happening. ie, why the first couple of experiments with Ranking seemed to work just fine, and then after that they would stop working. (rounding errors cause all significant digits to be quickly lost after a couple of Ranking changes)

I think the original data should be preserved in the original column, and a new column created in the data table, perhaps "Count (scaled)" or "Count-RANKED" or something like that. (If "Count" was the name of your original column.)

You could still tweak the specific rows, if necessary, and you would still have your original data.

Also, because the original data gets trashed, it can impact the Layout Algorithms. ... Which may, or may not, be a good thing.

--

PS: It would be convenient to be able to set the colors for a particular set in an edge Partitioning to be the same as the background color to "hide" a set of data temporarily. (If you can't get them to disappear by playing with the ranking and edge-size slider bar.) Unfortunately, when you do that, Gephi changes the color for those objects to a light gray -- making them even more visible than before.

What I've been trying to do is temporarily de-clutter a complex graph without deleting objects in order to highlight specific patterns of significance... It is not easy to do in the GUI yet.

[mbastian]

I'm sorry but I don't really get what is the issue you're describing with ranking. Could you be more precise?

You are trying to hide edges by changing their color. Though it might be a quick way it is not the "normal" way. Filters can do that for you. If you want manually to hide edges you could either:
- Create a new BOOLEAN column named 'hide' and edit the value in the data laboratory. With this true/false value, use a "Attribute Equal" filter to hide all 'false' values.
- Develop a plug-in, for example a 'Tool' (the icons in the toolbar), that requires of course to code. But the software is done to be extensible...we can't cover all use-cases by default (we are not a company, the software is free)

[rotten]

The problem I'm having with Ranking is that it changes the actual data in the column you run the ranking algorithm on. If you run it more than once (say to try different splines), you can lose all of your significant digits and it no longer has any effect.

Once you apply a Ranking, you have to reload your original data if you want to try a different one.

--

Changing the colors was something I was doing with partitioning. I could not easily set the color of any object to the same as the background color. Even when I set it to "white" (255,255,255), with a "white" background, the edges displayed as a dark gray.

I don't want to remove the edges of a particular dataset from the graph because I still want the force atlas layout to consider them when processing. I just want to _temporarily_ hide sets of them so that I have less clutter on the screen while I'm exploring the graph.

Sometimes it takes several hours for force atlas to run through my data until it stabilizes. I still like to look at it while it is running. I can adjust the edge size slider to wipe away the edges, but if I have two sets of edges with similar size ranges, it wipes away both.

My impression was filters would remove them from the graph.