Skip to content

A pagerank implementation in C++ able to handle very big graphs

Notifications You must be signed in to change notification settings

wenshan1/pagerank

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

An Open Source PageRank Implementation

This project provides an open source PageRank implementation. The implementation is a straightforward application of the algorithm description given in the American Mathematical Society's Feature Column How Google Finds Your Needle in the Web's Haystack, by David Austing. It can handle very big hyperlink graphs with millions of vertices and arcs.

Building

The project is written in standard C++ and can be built by running:

g++ -o pagerank pagerank.cpp table.cpp table.h

Usage

pagerank is invoked by

pagerank [OPTIONS] graph_file

where OPTIONS may be:

  • -t: if set, tracing is enabled. Tracing outputs all intermediate steps of the pagerank calculation algorithm and can be very voluminous.

  • -n: if set, the graph_file is in numeric format, that is it consists of lines of the form where and are integer vertex indices, starting with zero. If not set, the graph_file consists of lines of the form where and and vertex IDs, and will be interpreted as strings.

  • -a : the pagerank dumping factor; default is 0.85.

  • -c : the convergence criterion. The pagerank iterations will stop when two successive iterations have converged to less than or equal to this value. Default is 0.00001.

  • -s : the number of rows of the hyperlink matrix. This is not the maximum size; if the graph requires more rows, they will be allocated as necessary. If an approximate size is known beforehand, however, the necessary internal data structures can be pre-allocated for better performance.

  • -d : the delimited used to separate vector indices in the input graph file. Default is " => ".

Testing

Testing the implementation was carried out by comparing with pagerank calculations produced by Jung. The Java code that produced the calculations can be found in the java directory of the project. The test suites can be found at the test directory of the project, along with the test driver program pagerank_test.cpp, which is invoked by:

 pagerank_test <test_suite> 

The <test_suite> is a file containing in each line a filename, in the same directory, with an input graph. For an input graph foo.txt, the pagerank results against which we want to compare our implementation is found in foo-pr.txt. The result files are generated by the pagerank_calc.sh script.

The test driver is written in standard C++ and can be compiled with:

g++ -I../cpp -o pagerank_test pagerank_test.cpp ../cpp/table.cpp 

The graph test files were generated by the igraph R port using the R scripts in the test directory.

About

A pagerank implementation in C++ able to handle very big graphs

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published