Graphgrep uses an experimental HyperScan API call hs_compile_graph() to ingest automata pattern matchers using explicit Glushkov/Homogeneous/Position automata, rather than PCRE.
./graphgrep <graph file> <input byte stream file> <number of iterations>
If you have an ANML automata, vasim can convert it to the .graph format using the --graph flag. Graphs can also be created programmatically or by hand using the following format:
<number of nodes>
<node id> <character set bitlist> <start> <startDotStar> <report>
.... two examples below, one node per line
node_1_startds 0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001 0 1 0
node_2_accept 0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001 0 0 1
<from node id> <to node id> <to node id>....
.... two examples below connecting node_1_startds to node_2_accept, and the self-loop on node_2_accept, one per line
node_1_startds node_2_accept
node_2accept node_2_accept
Graphgrep currently requires my fork of HyperScan (jackwadden/hyperscan) to compile. You will have to clone and compile my fork and then edit the appropriate paths in the Makefile before running 'make'.