Questions and clarifications on errors, tolerances

[gkjohnson]

Related to #15 - I had a few questions about some of the errors that are thrown and resilience to input data so I can do my best to clean the data up before passing it into the library.

• The README states that there should be no duplicate points or points lying exactly on any constrained edges. But are there any tolerances that should be adhered to? Eg are points that are 1e-15 distance apart acceptable (I assume no)?
• Is a winding order guaranteed for the results or do I need to check the winding order before constructing a polygon?
• I'm occasionally seeing the error "Infinite loop: no further intersect after non-convex", but it's not exactly clear what this means. Is there a known graph configuration that will cause this error that I can try to avoid?
• Is the project resilient to adding the same constrained edge twice? What if it's added in the reverse order? eg adding [5, 2] and [2, 5]?

Thank you!

[kninnug]

Thank you for your questions, I'll answer them best I can:

• In principle, the tolerances should be exact to within the limits of JavaScript's number type (IEEE-754 double). Additionally, the library uses robust predicates for collinearity and intersection tests. So as long as the actual coordinate values are different, the points should not be seen as duplicates.
• Yes, as with Delaunator, the points of the resulting triangles are presented in counter-clockwise order. The same applies to the points on the hull.
• This error should not occur on well-formed Delaunator output, since it implies that the (hull of the) triangulation is not convex. If you can supply me an example dataset that throws this error, I would be very grateful :)
• Yes, the Constrainautor object maintains which edges are (already) constrained and constrainOne will return the half-edge id of the constrained edge when the caller tries to re-constrain the same points.

Of course, if you have any counter-examples to these claims, I would be glad to get them, so I can squash those bugs :)

[gkjohnson]

Thanks for the answers! When I get back to the project I was using this for I'll try to get some examples of the issues I was running into.