Vis roadmap (wrt. chap-links) ? about vis HOT 5 CLOSED

Well, there is no detailed roadmap. We just do maintenance/bugfixes for the chap-links-library, and add the features that the old Timeline had one by one to the new Timeline in vis.js. The new version has a much better architecture which allows to grow it further. The old Graph (line chart) will be merged in the new Timeline, this is now just another type of data to display on a time line.

If the only thing that you miss on the new Timeline is clustering, it may be the easy/doable to create this functionality yourself and add that to the new Timeline. You can prepare your data via some grouping/clustering function for different zoom levels (ie. group data together in clusters when zoomed out). Then attach an event listener to the rangechange event of the Timeline, calculate the current zoom level, and when the zoom level changed, replace the currently displayed dataset for the one corresponding to the new zoom level.

from vis.

If it's worth anything the clustering feature is something I could really use, having to look at alternatives again because of it not existing.

from vis.

If it's worth anything the clustering feature is something I could really use, having to look at alternatives again because of it not existing.

same for us. But don't expect it within a few months. Unless you implement it right away and make a pull request to share it with us ;)

from vis.

I am looking more seriously into this. I have a question regarding the integration of such a feature. Given that the code is basically the same as the existing Stack component (since we need to detect overlapping items, and represent them appropriately), what would be the most appropriate/generic way to implement clustering? Should we try to leverage the Stack.js code to handle both stacking and clustering, or should we leave the Stacking alone and do the clustering at the data level?

Based on your initial suggestion, I have the idea of providing a "transform" property, consisting in a list of operations that will handle transformations to the data based on timeline parameters such as zoom level and width. We can bundle standard transforms such as clustering, filtering... It can be implemented at the DataSet level, but then I need to find a way to pass view-specific parameters (timeline zoom/width for instance) so that the transforms are done according to the view.
Any feelings/comments about this?

from vis.

There are a couple of discussions on the chap-links-library discussing various clustering solutions. I think there are different ways, and in the end we may up supporting a couple of them. Few ideas:

• clustering on a data level, purely taking into account timestamps and a given time window.
• not doing clustering by yourself, but offering an easy interface for a developer to define different datasets for different zoomlevels.
• Clustering by really reckoning with the actual size of items on screen, making everything fit exactly in the visible window. This will require integration with Stack.js.

So, depending on your needs, you could start implementing one of these.

from vis.