IntelligentSoftwareSystems/Galois

How to convert the graphalytics benchmark's dataset to gr file?

photoszzt opened this issue · 2 comments

The dataset is from here: https://graphalytics.org/datasets
The data has a file that encodes edge(.e) with a tuple and a file that contains vertex(.v).

I try the following:

./build/bin/graph-convert --edgelist2gr --edgeType=void inputs/graph
alytics_xs/example-undirected/example-undirected.e inputs/graphalytics_xs/example-undirected/example-undirected.e.gr
Graph type: void
InGraph : |V| = 11, |E| = 12
OutGraph: |V| = 11, |E| = 12
STAT_TYPE, REGION, CATEGORY, TOTAL_TYPE, TOTAL

On the website, the example-undirected graph has 9 vertexes and 12 edges. I don't know how to add the vertex list. It seems the input only accept one file.

By the way, what's the gr format? Is it a form of Compressed Sparse Row (CSR) format stored on disk?

I'm sorry for the late response.

  1. I took a look at example-undirected: it seems it's actually working as intended.
    You don't actually need the vertex list since the converter will determine the # of vertices
    from the edges list (max index is # of vertices: in this case it's 10, and it's 0-indexed, so vertices
    ends up 11).

To keep the edge types as well, you need to specify the edgeType option to the converter
as "float". Use -help to get the exact command.

  1. Yes, the gr format is just a binary CSR on disk with a header.