Fig. 1
From: A graph extension of the positional Burrows–Wheeler transform and its applications
An illustration of the \(B_{1}[]\) array for a single side numbered 1. (Note that a similar, reverse view could be constructed for the \(B_2[]\) array and the opposite orientations of all the thread orientations shown here, but it is omitted for clarity). The central rectangle represents a node, and the pairs of solid lines on either side delimit edges attached to either the left or right side of the node, respectively. These edges connect the node to other parts of the graph, which have been elided for clarity. The dashed lines within the edges represent thread orientations traveling along each edge in a conserved order, while the solid lines with triangles at the ends within the displayed node represent thread orientations as they cross over one another within the node. The triangles themselves represent “terminals”, which connect to the corresponding dashed lines within the edges, and which are wired together within the node in a configuration determined by the \(B_{1}[]\) array. Thread orientations entering this node by visiting side 1 may enter their next nodes on sides 3, 5, or 7, and these labels are displayed near the edges leaving the right side of the diagram. (Note that we are following a convention where nodes’ left sides are assigned odd numbers, and nodes’ right sides are assigned even numbers). The \(B_1[]\) array records, for each thread orientation entering through side 1, the side on which it enters its next node. This determines through which of the available edges it should leave the current node. Because threads tend to be similar to each other, their orientations are likely to run in “ribbons” of multiple thread orientations that both enter and leave together. These ribbons cause the \(B_s[]\) arrays to contain runs of identical values, which may be compressed.