Example code for a simple scenario, input List of Points --> output List of Triangles (List of 3 point array)?

[istinnstudio]

Hi, I have only a general view about this library so before going deep I would like to know if it is possible to just use the core implementation in order to simply triangulate path2D polygons (or Lists of points) and get a list of the triangles, much like those libs do:
https://github.com/jdiemke/delaunay-triangulator
https://github.com/orbisgis/poly2tri.java

Paths or List of points as input, list of triangles (3 point Lists) as output, those can then be converted to paths again and used as Graphics2D objects. The use case is art projects, not geographic information systems, so objectives are different, the coordinate precision remains critical though, as "double".
Is there a code example that can be provided here in order to cover this simple scenario, input List of points--> output List of Triangles ?

Can we also tweak the result by using other data as "constrains" additional "list of points" or polygons?
thanks!

[gwlucastrig]

If you have a Path2D object, extracting the points is a separate task. But here's an example of how it would work if you had the points to start with.

There is more explanation on the Tinfour wiki article Three Easy Iterators

  public static void main(String[] args) {
    // make a polygon, just as an example
    int s = 1; // length of side
    int n = 6; // number of sides
    IncrementalTin tin = new IncrementalTin(1.0);
    for(int i=0; i<n; i++){
      double theta = i*2*Math.PI/n;
      double x = s*Math.cos(theta);
      double y = s*Math.sin(theta);
      Vertex v = new Vertex(x, y, 1, i);
      tin.add(v);
    }
   
    // get the triangles
    for(SimpleTriangle triangle:  tin.triangles()){
      System.out.println(""+triangle);
    }
  }

[gwlucastrig]

A Delaunay triangulation always creates a convex enclosing polygon. So if your art object is non-convex, the code example will produce some triangles that are exterior to the art object. If this is a problem, you will have to define constraint objects, which is a somewhat more complicated problem.

To learn more about how that part of the Tinfour API works, see: Using Polygon Based Constraints

If you need help with that, I can probably lend some assistance.

[istinnstudio]

Thank you!! very usefull reply, I have used the library with success in thousands of polygons. This example was very informative in order to quickly see the sequence of functions.
And yes, the case with some non-convex polygons is present, the external triangles effect is nice and surely usable, but I will try to find the way not to create them in those spaces. I will see if I can find the way in ... Polygon Based Constraints
Can I use the original path in order to create a Constraint for those cases?

[gwlucastrig]

The geometry from the original path may be used provided it is not self intersecting. Also, it must be oriented counterclockwise. The Java Path2D class can contain a few objects that are not supported such as curved arcs. You can use some of the techniques in the wiki article to know if a triangle is inside your polygon or outside

…

On Mon, Feb 6, 2023, 3:04 PM istinnstudio ***@***.***> wrote: Thank you!! very usefull reply, I have used the library with success in thousands of polygons. This example was very informative in order to quickly see the sequence of functions. And yes, the case with some non-convex polygons is present, the external triangles effect is nice and surely usable, but I will try to find the way not to create them in those spaces. I will see if I can find the way in ... Polygon Based Constraints <https://github.com/gwlucastrig/Tinfour/wiki/Tutorial-Using-Polygon-Based-Constraints> Can I use the original path in order to create a Constraint for those cases? — Reply to this email directly, view it on GitHub <#95 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEWJDYK46C5EJRYYY7OOVOLWWFKMJANCNFSM6AAAAAAUSEB4PY> . You are receiving this because you commented.Message ID: ***@***.***>

[istinnstudio]

Thanks for the support, it seems that for now I cannot find a way to remove those exterior triangles. But, if there would be an example based on the basic code above (with convex and non convex source) that would be great. As a final thought, as far as I can see, there could be a method that returns boolean and if called can "flag/mark" those triangles at the very begining of the tin creation, so all of them can be selected and used/removed afterwards without further conversions or "contains" calculations. Just thinking, not sure if it is a good idea or have application in other cases.

[istinnstudio]

Also, PGS does live in native Java! You can use it as a maven library alongside Processing to code in Java (as I generally do).
@micycle1

I will defenately try that out, I generally avoid processing "extends PApplet" class because my project is based mainly on all java graphics2D native methods for various reasons and shapes are being formed in Path2D.Double. Although I can get some PImage "bufferedimages" from processing to work with, I have never tried to work with i/o vector data, points, shapes etc.

So when you say : "The focus of the library is on visualisation rather than providing underlying data structures. To this end all methods in the library are static and most of them take in and return PShapes or PVectors."

you meen that if I can code some simple convertions between PShape <--> Path2D and PVector <--> Point2D (like I do for the tinfour for Vertex), then we can use the library as we would do in processing ide or extending the processing.org PApplet class? Is this possible? What about the "float" precision of the processing.org coordinates?

[micycle1]

The conversions are already coded, so you needn't even work in Processing at all -- you can work with what you have and convert awt shapes, process them with PGS, then convert back to display them.

[istinnstudio]

Nice!! I am curious how the float precision of Processing Objects will "react" with double precision Path2Ds, and if this "downgrade" will affect visual appearence in high magnifications or very small areas... (I imagine hybrids that both co-exist keeping original shape boundaries double, filling them inside with pgs float elements) nevertheless, so many oportunities are being opened and this is great.

[micycle1]

The precision loss of double->float conversion is negligible (think at most ~1e-7?), particularly in the context of computer graphics.

[istinnstudio]

Issues with complex drawings with thousands of polygons do exist. But I will have to test in order to find out if those limits can be reached, e.g. in svg exports. A related post can be found here : https://blog.generativedarkroom.com/why-i-switched-from-processing-to-openrndr-b50c931d21ae