[igraph] About bliss isomorphism testing and potential bug in igraph_isomorphic_bliss
Dear All, I have a few questions about C/igraph and isomorphism testing using bliss. According to the documentation, igraph uses bliss 0.35 http://igraph.org/c/doc/igraph-Isomorphism.html#idm470920858320 * Is this accurate? If yes, is there a specific reason why igraph doesn't upgrade to more recent bliss? Currently 0.50 and 0.72 are available. 0.50 seems to have functionality that 0.72 doesn't. Maybe the same is the case with 0.35? The R/igraph documentation says that bliss doesn't support directed graphs: http://igraph.org/r/doc/isomorphic.html The C/igraph documentation makes no such mention and the canonical_permutation function does return a result for directed graphs. * Does bliss support directed graphs in igraph? All igraph_bliss functions return an igraph_bliss_info_t which has the group_size member. * Do I need to _always_ free(group_size), even if I ran some other function than igraph_automorphisms()? Finally, the bliss documentation says: http://www.tcs.hut.fi/Software/bliss/doxy/classbliss_1_1Graph.html#08da370e34106cd7db479eca7c7375cc "The selected splitting heuristics affects the computed canonical labelings; therefore, if you want to compare whether two graphs are isomorphic by computing and comparing (for equality) their canonical versions, be sure to use the same splitting heuristics for both graphs." But igraph_isomorphic_bliss takes a _separate_ splitting heuristic argument for each of the two input graphs. It would seem that setting different splitting heuristics may cause the function to return a wrong result. * Is it a bug that different splitting heuristics are allowed? Or is it done intentionally? Szabolcs ___ igraph-help mailing list igraph-help@nongnu.org https://lists.nongnu.org/mailman/listinfo/igraph-help
Re: [igraph] How bad an idea is to access the igraph_vector_t and igraph_matrix_t structs directly?
I'm sorry, I realize I made a mistake and igraph does store matrices in column major order. On 30 August 2015 at 21:12, Szabolcs Horvát wrote: > Dear All, > > The igraph_vector_t and igraph_matrix_t structs don't seem to be > documented. Does this mean that they are not meant to be used by end > users and I should stick to the documented API instead of accessing > the structs directly? > > For my use case, working with the structs is just more convenient, > especially for matrices. I use C++ style being and end iterators > which seems to be stored int he struct and stor_begin and end. I need > to copy matrices as fast as possible in row-major order, which seems > to be the internal storage format, while matrix_copy_to() uses > column-major format. > > So, to sum up: how stable is this interface and is it very risky is I > use it directly? > > Szabolcs ___ igraph-help mailing list igraph-help@nongnu.org https://lists.nongnu.org/mailman/listinfo/igraph-help
[igraph] How bad an idea is to access the igraph_vector_t and igraph_matrix_t structs directly?
Dear All, The igraph_vector_t and igraph_matrix_t structs don't seem to be documented. Does this mean that they are not meant to be used by end users and I should stick to the documented API instead of accessing the structs directly? For my use case, working with the structs is just more convenient, especially for matrices. I use C++ style being and end iterators which seems to be stored int he struct and stor_begin and end. I need to copy matrices as fast as possible in row-major order, which seems to be the internal storage format, while matrix_copy_to() uses column-major format. So, to sum up: how stable is this interface and is it very risky is I use it directly? Szabolcs ___ igraph-help mailing list igraph-help@nongnu.org https://lists.nongnu.org/mailman/listinfo/igraph-help
