Opportunity for performance gains via d3-force-reuse?

[Nantris]

Is it possible that the code in this repo might still bring performance advantages, and is there any interest in integrating it into d3-force?

https://github.com/twosixlabs/d3-force-reuse

[loleg]

First, I would ask what the most compelling reason today would be to improve performance of this library - i.e. which use cases demand it, which are not better served by other force-directed graphs (e.g. PIXI.js, THREE.js)?

Second, I would check how the "reuse" approach stands in comparison to others, like making the situation more parallel using workers as suggested in #112 and demonstrated by @dileban/nbody-simulation. The other day, an excellent overview of n-body algorithms was posted in Efficient Numerical Methods for N-Body Simulations with Modern Computational Techniques by Nayer Sharar.

Finally, I would suggest investing a bit of effort into further testing the d3-force-reuse library, making sure there are no edge cases that are affected by this technique. If the improvements are reliable, it should be the default method.

The question of merging was also raised in twosixlabs/d3-force-reuse#2