Create the Transaction Graph

1. Run the following query in the psql shell to write the edges file to disk.

   \COPY (SELECT keyimage_id, output_id FROM xmr_bigraph_edges) TO '/tmp/fork-ringct-edges-2530000.txt' WITH DELIMITER ' ';
   

2. Each row in the fork-ringct-edges-2530000.txt file represents an edge. The keyimage_id and output_id values that represent the edge are index values from the respective PostgreSQL tables. These index values do not form a contiguous range and can have gaps. But the sparse graph representations we will use work better without gaps in the index ranges.
   • Copy the fork-ringct-edges-2530000.txt file to the scripts/monero directory.
   • Compile the create_csparse_edges.cpp file located in the scripts/monero directory. Run it with the block height as argument.

     cd scripts/monero
     g++ -O2 create_csparse_edges.cpp 
     ./a.out 2530000 fork-ringct
     

     The output should look like the following.

     Reading edge file
     Finished reading edge file
     Number of key images: 40351733
     Number of outputs: 45805316
     Number of vertices: 86157049
     Number of edges: 409132035
     Creating keyimage index map
     Finished creating keyimage index map
     Creating output index map
     Finished creating output index map
     Adding edges to graph
     Finished adding edges to graph
     

   • Three output files are created.

     • fork-ringct-csparse-edges-2530000.txt

       The edges in this file are represented as pairs of indices starting from 0 followed by a 1. This is the CSparse format for representing sparse matrices. The 1 corresponds to the value of the matrix entry. In our case, we use this format to represent sparse bipartite graphs.

     • fork-ringct-index-keyimageid-map-2530000.txt

       This file has pairs of indices and key image IDs per line. It is to translate the result of the transaction graph analysis back to the key image ID space.

     • fork-ringct-index-outputid-map-2530000.txt

       This file has pairs of indices and output IDs per line. It is to translate the result of the transaction graph analysis back to the output ID space.