Possible to "cluster" or group nodes based on keywords or tags (differ from the normal "nearest neighbors linked nodes) One thing that would definitely be nice would be to be able to see the yaml title or H1 rather than the filename for a node, my filenames are ids which is uninformative without opening the file.Ĭolorize Nodes based on "type" or other Yaml keypair Maybe it would be interesting if I could highlight a node and have all the ones that it links to light up, optionally two or three levels deep?Īlso I'm not sure what the purpose of zooming in is? I had assumed the spaces between nodes would get bigger so I could see the connections, or something like that, but when I zoom in with the mousewheel I get a screenful of densely packed big dots! Screenshot Have I missed a setting or way to use it?
ZETTLR GRAPH VIEW PLUS
It does tell me that I've only really got one big cluster of most of my ~2.5k files, plus a small handful of isolates and isolated mini-clusters, so maybe that's why it feels less informative. However, even though I was thinking it probably wouldn't be useful I guess I expected something a little more informative than I am getting, see screenshot. I've gone back and forth over the last couple of years on whether a graph view would be useful (as opposed to pretty), so I'm not the best person to give feedback/suggest features as I was coming down on the side of "maybe pretty but that's all". The graph building process still takes a lot of time, depending on how many links you have, there is lots of room for optimization.īeta Was this translation helpful? Give feedback.Links overlap each other and in general the simulation is somewhat awkward (I couldn't find any solutions for this yet).Currently the graph will rebuild itself even if nothing in the link structure has changed (i.e.Dragging the network graph does not yet work.Clicking on a vertex should highlight the path from its parent to all leaf nodesĪlso, there are a few known issues that I will be gradually improving, so again: Please do not open an issue for these as well!.Possibly integrate a better layouting mechanism such as Fruchterman-Reingold or MDS-based layouts (e.g., Kamada/Kawai).Different color shading of the links between nodes based on them being either asymmetrical or mutual ( MAN criterion).Weighing of edges, for example utilizing edge betweenness centrality.Implement HITS (Hubs/Authority detection algorithm).Currently the community detection uses the weak criterion for clustering components I plan to also add the strong community detection criterion (which will identify different components even if they are connected, if there is one link which is asymmetrical, and not mutual).Add arrow heads to the links between nodes to visualize the directionality of the arcs.Highlight specific nodes without completely hiding the others.Center the view on specific nodes (implemented).Filter nodes by path/folder/title/tags (implemented with path, not yet tags).Allow vertices to be scaled according to various node centrality measures (eigenvector centrality, betweenness centrality, degree centrality).I'll also consult faculty at my institute to make the graph view scientifically correct. I am planning on playing around with a lot of different ways of analyzing and visualizing networks since I have to do that for my PhD as well. There are a few features I already have planned, please refrain from opening issues for these.
Please utilize the tree-like structure of comments in this thread to try to keep it a little bit ordered :) Already Planned This is meant to prevent the issue tracker from flooding with numerous different feature requests. So one word of caution: If you would like to have a feature for the graph view, please discuss it here and DO NOT OPEN AN ISSUE for this.
That being said: In the future I plan on adding more features depending on what works, and what doesn't. As I made clear in the changelog, the current implementation is still in its infancy. Today, Zettlr 2.2.2 shipped with the new graph view.
0 kommentar(er)
