Alexander van der Grinten

Results 119 comments of Alexander van der Grinten

The intent of that option is disable overly expensive assertions. I guess `assert` expresses this intend more clearly but you are right that NETWORKIT_SANITY_CHECKS + NDEBUG does not really make...

NetworKit should be some orders of magnitude faster for combinatorial algorithms (such as Betweenness, Closeness, community detection algorithms, distance algorithms, etc) since networkx implements them in pure Python (as far...

Thank you for the report. One temporary solution is to install NetworKit via WSL. But of course, the compiler should be detected correctly even on native Windows.

Good objection. I did not consider `none`. I will review the checks in `Graph.cpp`. It might make sense to add an assertion that prevents `none` from being added to the...

I took a quick look at the checks. `none` is mainly used as a return value in functions like `indexInInEdgeArray`. There are one or two checks that test for `none`...

This PR needs some comments on the general approach and whether it should be adapted or not. Regarding the motivation for this PR: In an application, I had to fit...

Instead of changing parallelForNodes(), I would prefer to update the centrality computations to properly take deleted nodes into account. Letting everything that uses parallelForNodes() deal with node indices that can...

I plan to prepare a fix for this issue for the next NetworKit release. As discussed above I do not plan to change parallelForNodes() but integrate the fix into the...

This should probably be re-evaluted and moved to a new issue ("Deleted nodes are not handled correctly in all algorithms" or so).

AFAICS, the two functions are unrelated for MSI-X: one enables MSIs in general, the other one installs one MSI vector (there can be multiple vectors for MSI-X). For plain MSI,...