Fig. 3

Depiction of the internal data structure: Atoms are represented as nodes and bonds as edges. To the nodes, a number of “node attributes” is attached which are either used in the canonicalization (blue) or in graphical output (magenta). Generally, the edges could also be assigned “edge attributes” but these are not needed for canonicalization. Furthermore, nodes and edges can also be assigned user-defined “custom attributes”, which are also not used in the canonicalization, but might be useful to carry along data, for example for machine-learning applications