check for mass value of zero on vis.js Network nodes.

[pontusvision]

I was updating an old library to the latest vis.js network, when I spotted quite an odd behaviour, where only one node out of hundreds would appear. After a fair amount of debugging, I noticed that I was accidentally setting the mass some of the nodes to zero. This was causing havoc, with a NaN eventually as the x and y coords.

I've created the following jsfiddle to demonstrate it:
http://jsfiddle.net/1y8L119L/

Setting the mass to 1 or higher solves the issue; however, the code used to work with a mass of 0 back in version 4.3.0.

Please add the extra safety check to prevent the pain I have just had to go through on other distracted silly programmers that set the mass to zero.

[wimrijnders]

There's two years between 4.3.0 and current 4.20.0. I honestly tried to dig back through time to see what changed, but I can't manage it. So I'll skip trying to clarify what changed in the meantime to make this happen.

But really I have to agree with you. The value zero should be tested in some way and reported. I'll be paranoid and have a look if negative values should be disqualified as well. Thanks for reporting.

---

Background info:

Then, as now, barnesHut was the default physics solver, which in the docs comes with the warning for the mass field:

Values lower than 1 are not recommended.

Methinks that zero qualifies.

[wimrijnders]

It turns out that setting negative mass is a really bad idea. The nodes make a sincere attempt at escaping and disappear into the distance, never to stop. I'm inclined to block negative mass as well.

[wimrijnders]

The question now is how to handle this. Should vis.js just plain throw an error when this happens? Or perhaps a warning and then reset the masses (to 1 would be most obvious)?

Update: I'm going for warning and reset to 1.

[wimrijnders]

I've added the checks, they should be in the next release. Thanks for reporting!

Please close the issue if you're happy with this.