frankmcsherry / dataflow-join Goto Github PK

may share links to similar functionality code but in python?

I'm trying out a few examples after reading your blog post on tracking motifs, and unless I'm doing something wrong, it looks like the number of motifs found by examples/motif.rs may be underestimated.

Using the same livejournal dataset from the blog, and with a "single directed edge" motif, I get results that are a lot smaller than the expected (which I believe should match the number of edges in the graph, or at least be somewhat close to it):

$ wc -l /data/soc-LiveJournal1.txt
 68993777 /data/soc-LiveJournal1.txt
$ cargo run --release --example motif -- 1 0 1 /data/soc-LiveJournal1.txt 68000000 1000 inspect
(...)
elapsed: Duration { secs: 22, nanos: 760898251 }	total motifs at this process: 993777

$ wc -l /data/soc-LiveJournal1-1000.txt
    1000 /data/soc-LiveJournal1-1000.txt
$ cargo run --release --example motif -- 1 0 1 /data/soc-LiveJournal1-1000.txt 1000 1000 inspect
(...)
elapsed: Duration { secs: 0, nanos: 1657901 }	total motifs at this process: 4

In the first case, for the livejournal graph with close to 69 million edges, the example found less than 1 million "single directed edge" motifs. For a chunk of the same graph which contains only the first 1k edges, it found only 4 instances of that same motif.

The reason why I started looking at these simple "single directed edge" motifs was because I was trying to debug a feed-forward-loop and other slightly more complex motifs. In particular, I have a few synthetic graphs, which I know contain the motifs I'm looking for, but that example/motif.rs fails to find. Please advise, thanks! --Joana