Problem:
Rows in the output file for paths have a column "KSP index" corresponding to the order the path was found. Rows in the ranked edges file have a corresponding column, referring to the index of the first path the edge was seen in. At first blush, one might be tempted to treat this as a rank for paths/edges.
However, the KSP index of these paths thus depends on implementation-dependent priorities that determine the orders paths of equal weight are found in, rather than the presumed path-weight ranking criteria. Two paths with the same weight will not receive the same KSP index.
This serves to make the ranked-edges file and to some extent the output file for paths counterintuitive, as they discriminate between edges/paths on arbitrary criteria.
Expected behavior:
The output file listing paths should give the same KSP index to paths with the same rank
The output file listing edges should otherwise remain unchanged, with the KSP index of the edge referring to the updated, implementation-agnostic index of the path
Reproduce:
This can be seen in our sample output files
This is not a bug, but I do think it violates the principle of least astonishment.
Problem: Rows in the output file for paths have a column "KSP index" corresponding to the order the path was found. Rows in the ranked edges file have a corresponding column, referring to the index of the first path the edge was seen in. At first blush, one might be tempted to treat this as a rank for paths/edges.
However, the KSP index of these paths thus depends on implementation-dependent priorities that determine the orders paths of equal weight are found in, rather than the presumed path-weight ranking criteria. Two paths with the same weight will not receive the same KSP index.
This serves to make the ranked-edges file and to some extent the output file for paths counterintuitive, as they discriminate between edges/paths on arbitrary criteria.
Expected behavior:
Reproduce: This can be seen in our sample output files
This is not a bug, but I do think it violates the principle of least astonishment.