Floyd-Warshall algorithm #98
Conversation
Corrected trailling slashes
- relative railway example - corrected png files
- Added test cases and docs
Added one test
Corrected doc file
- algorithm - test - documentation
Codecov ReportAttention:
... and 2 files with indirect coverage changes 📢 Thoughts on this report? Let us know!. |
| template <typename V, typename E, graph_type T, typename WEIGHT_T> | ||
| std::vector<std::vector<WEIGHT_T>> floyd_warshall_shortest_paths( | ||
| const graph<V, E, T>& graph) { | ||
| WEIGHT_T ZERO{}; | ||
| std::size_t n = graph.vertex_count(); | ||
| auto INF = std::numeric_limits<WEIGHT_T>::max(); | ||
| auto NO_PATH = std::numeric_limits<WEIGHT_T>::min(); |
There was a problem hiding this comment.
Regarding using std::numeric_limits<WEIGHT_T>::min() for vertex pairs between which we cannot find a path (i.e. vertices in negative weight cycles), consider the case where the WEIGHT_T is an unsigned type.
In that case, the caller cannot distinguish between a NO_PATH and a true distance of 0. However, I do not see an elegant way of handling this. What do you think about removing the negative cycle detection, as it is also documented on the interface that the algorithm does not work in this case.
There was a problem hiding this comment.
Oh, I didn't see this one. My bad. I'll fix it right away. Agreed, we could move it to a different PR as Floyd-Warshall negative cycle detection so results will be less confusing. And in this case, NO_PATH will no longer be needed.
I look forward to your next comments about PR. :)
Added Floyd-Warshall alogorithm, tests and docs.