Problems with Numerical Indexing

Idea of numerical indexing is favourable because one can have precise control over the boundaries of shapes and intuitively is independent of angles of geometry – in contrast to geometrical selection (based angles of edges and similar constructions). In practice it has turned out to be very hard to track numeric indexing, for instance even peel range parameter is altered to extend to one more LE (Logical Edge). Moreover as operations are stacked up geometric complexity rises and numeric indexing becomes increasingly chaotic. As a results a simpler indexing system was devised.


The idea is to still have number-based indexing, but to reduce number of labels to just two: “0” and “1”. The reason for two labels is that they naturally map to park layout where “0” corresponds to boundary edges and “1” marks the inner edges – with respect to the parent shape (see picture below).Though it is possible to have yet even simpler ‘unary’ indexing in range [0, 1) however in park designs it is important to distinguish between inner and outer boundaries, so arguably “01” indexing is ‘sufficiently minimal’ indexing in context.

01indexing behaviour for partitioning rules and peel
01indexing behaviour for partitioning rules and peel


In some cases, for instance peel rule, there are two variants in what is labelled “1”:

  1. Edges spanned by inner vertices, or
  2. Edges where one of the vertices is inner and one outer, as in the shaped part of the image above.

Decision was to use the second approach where “0” indexing requires edges to be spanned by the boundary vertices only. Later a rule flags can be added to target exclusively inner edges.



The shape boundary can not only be partitioned but modified for instance in place rule – in such case newly inserted vertices has to be relabelled (see picture). The following cases are handled:

  1. Completely inner – trivial, labelled the same way regardless of direction.
  2. Between two ranges with different labels – currently a simple solution is used where an unlabelled range are covered by the whichever label occurs first (“0” as shown on the picture above).
  3. Data boundary (first index) occurring within the unlabelled range. This problem is handled by first rotating the whole polygon to start at the first zero-labelled edge.


Peel rule can result in splitting of the substrate part (image1 below) in two (or more) subparts when extruded part is large enough. Similar fragmentation can be caused by insertion of a placement shape (image2 below).