4.3. graph-canon-run

The graph-canon-run program can be used to run a specific graph canonicalistion program on a batch of graphs (e.g., downloaded with download-graph-collections).

4.3.1. Graph Database

The program indexes graph instances from a specific folder:

GRAPH_CANON_DATA_DIR

If this environment variable is defined, it is assumed to be the path to a folder containing a graph database. If it is not defined the database folder is assumed to be the relative path graphs/.

A graph database is simply a folder with a specific structure: the root contains folders representing “packages”. For example, download_graph_collections program will download graphs from http://pallini.di.uniroma1.it/Graphs.html and partition them into the packages bliss, conauto, saucy, random, and misc, corresponding to the different sources benchmark graphs. Each package subfolder is then assumed to have folder representing “collections” of graphs. For example, the bliss package will have the collections ag, cfi, cmz, etc. The names of collections are assumed to be uniq, even across packages. Each collection subfolder finally contains text files with one graph each. For example, the file bliss/ag/ag2-2 would be a file with a graph.

4.3.2. Listing Mode

--list-collections

Print the list of available collections of benchmark graphs and then exit. Each collection is printed on a separate line as <package>/<collection>.

--list-instances [package|collection]...

Print the list of instances contained in the given packages or collections. Each instances is printed on a separate line as <package>/<collection>/<instance>.

--get-instance-size [package] [collection] [instance]

Print the number of vertices in the given instance and then exit. This assumes the graph is stored in DIMACS format, with the first line specifying the graph size (i.e., the p edge <n> <m> line is the very first).

4.3.3. Execution Mode

--args [arg]...

Additional arguments passed through to the canonicalization program.

-t [seconds], --timeout [seconds]

Time before the canonicalization is killed. However, if --memcheck or --debug is given, the time limit is not in effect.

-e [executable], --exe [executable]

The executable to use for canonicalization. It defaults to graph-canon (i.e., the graph_canon program, as found through normal shell lookup.

--memcheck

--debug

--profile

Options passed through to the executable.

[task]...

Each positional argument is assumed to be a “task”, which must be the name of either a package, a collection, or the path to an instance. The executable will in turn be invoked on each instance specified. The special meta-task all can be used to specify all instances in the graph database.

Each invocation of the canonicalization executable will be done with -f <full path to graph instance> for specifying the current input graph- Additionally the arguments --id "GCDResult <package>    <collection>    <instance>", are passed, meant to allow the executable to prefix each data line with this string for easier tabulation of results.