yamafaktory / hypergraph Goto Github PK

The integration test in tests/test.rs seems to create an invalid dot file (which is normally just printed to the void and not actually used). The output is:

digraph {
    edge [penwidth=0.5, arrowhead=normal, arrowsize=0.5, fontsize=8.0];
    node [color=gray20, fontsize=8.0, fontcolor=white, style=filled, shape=circle];
    rankdir=LR;

    0 [label="Vertex {\l    name: \"d\",\l}\l"];
    1 [label="Vertex {\l    name: \"b\",\l}\l"];
    2 [label="Vertex {\l    name: \"c\",\l}\l"];


     [color="#0c8946", fontcolor="#0c8946", label="\"yup\"\l"];
     [color="#5309a3", fontcolor="#5309a3", label="\"foo_\"\l"];

    0 -> 2 [color="#5b7a07", fontcolor="#5b7a07", label="\"bar\"\l"];
}

It seems to be outputting two edges with connection zero nodes, so just the attributes are being printed? When I run this through dot it errors (because it expected "A -> B" befre the "[color=..."). Is this an error in the dot writing code which is producing vertices without members?

Are there any databases that are supported here? I would like to use this in production.

Any aspirations to add PCSR, maybe as an optional graph representation?

From the article:
"...PCSR was orders of magnitude faster for inserts and updates than CSR
and adjacency list while maintaining similar graph traversal times."

http://supertech.csail.mit.edu/papers/WheatmanXu18.pdf

PCSR Code:
https://github.com/wheatman/Packed-Compressed-Sparse-Row

IMHO, space (memory) is easier to solve than the time it takes to generate a graph from a large dataset.

I'm generating currently a number of hyper graphs that usually reach between half a million and 2 million vertexes & edges. My current worst case runtime is in the ballpark of 15 minutes.
These graphs are based on 10% data sample size, the non-sampled data would be more in the ballpark of 30 to 50 million vertices and edges per graph hence the insert time becomes a bit more critical.

Thoughts?

Would you be averse to me refactoring your graphviz code to use my new dot-writer library? Largely for my own selfish reasons (so I can put it through its paces and put it out in the world a bit), but arguably it could make your code more readable, strongly typed and terser?

Happy to do the pull request myself of course (I might have to anyway as it might be missing a function here or there that your dot output needs), and equally understand if you'd like to minimize dependancies / leave things as they are.