WebGraph is a framework to study the web graph. It provides simple ways to manage very large graphs, exploiting modern compression techniques. More precisely, it is currently made of:

  1. A set of simple codes, called ζ codes, which are particularly suitable for storing web graphs (or, in general, integers with a power-law distribution in a certain exponent range).
  2. Algorithms for compressing web graphs that exploit gap compression and differential compression (à la LINK), intervalisation and ζ codes to provide a high compression ratio (see our datasets). The algorithms are controlled by several parameters, which provide different tradeoffs between access speed and compression ratio.
  3. Algorithms for accessing a compressed graph without actually decompressing it, using lazy techniques that delay the decompression until it is actually necessary.
  4. This package, providing a complete, documented implementation of the algorithms above in Java. It is free software distributed under the GNU General Public License.
  5. Data sets for very large graph (e.g., a billion of links). These are either gathered from public sources (such as WebBase), or gathered by UbiCrawler.

In the end, with WebGraph you can access and analyse very large web graphs. Using WebGraph is as easy as installing a few jar files and downloading a data set. This makes studying phenomena such as PageRank, distribution of graph properties of the web graph, etc., very easy.

You are welcome to use and improve WebGraph! If you find our software useful for your research, please quote our paper “The WebGraph Framework I: Compression Techniques”, by Paolo Boldi and Sebastiano Vigna, in Proc. of the Thirteenth World–Wide Web Conference, pages 595−601, 2004, ACM Press.

Looking around

Warning: WebGraph 2+ is not fully compatible with previous versions and requires some minor code refactoring. Please refer to the documentation of {@link it.unimi.dsi.webgraph.ImmutableGraph}.

For in-depth information on the Webgraph framework, you should have a look at its home page, where you can find some papers about the compression techniques it uses. Datasets are available at the LAW web site.

The classes of interest for the casual Webgraph user are {@link it.unimi.dsi.webgraph.ImmutableGraph}, which specifies the access methods for an immutable graph, {@link it.unimi.dsi.webgraph.BVGraph}, which allow to retrieve or recompress a graph stored in the format described in The WebGraph Framework I: Compression Techniques, and {@link it.unimi.dsi.webgraph.Transform}, which provides several ways to transform an {@link it.unimi.dsi.webgraph.ImmutableGraph}.

If you plan on building your graphs dynamically, the class {@link it.unimi.dsi.webgraph.ArrayListMutableGraph} makes it possible to create incrementally a graph and then extract an {@linkplain it.unimi.dsi.webgraph.ArrayListMutableGraph#immutableView() immutable view}.

The package {@link it.unimi.dsi.webgraph.examples} contains useful examples that show how to access sequentially and randomly an immutable graph.

Importing your data

If you want to import your own data into WebGraph, you must write an implementation of {@link it.unimi.dsi.webgraph.ImmutableGraph} that exposes your data. A simple example is given in {@link it.unimi.dsi.webgraph.examples.IntegerListImmutableGraph}, a stub class exposing a simple, noncompressed binary format as an {@link it.unimi.dsi.webgraph.ImmutableGraph}. Once your data is exposed in that way, you can get a compressed version using the store() method of your class of interest. Often, there is a main method (see, e.g., {@link it.unimi.dsi.webgraph.BVGraph}) that will load your class and invoke store() for you.

As an alternative, the class {@link it.unimi.dsi.webgraph.ASCIIGraph} can be used to read graphs specified in a very simple ASCII format. The class implements {@link it.unimi.dsi.webgraph.ASCIIGraph#loadOnce(java.io.InputStream)} so that the file can be just piped into a class offering a main method that supports loadOnce() (e.g., {@link it.unimi.dsi.webgraph.BVGraph}). You can also generate a graph in ASCII format and read it using {@link it.unimi.dsi.webgraph.ASCIIGraph#loadOffline(CharSequence)}—the graph will not be loaded into main memory.

{@link it.unimi.dsi.webgraph.ASCIIGraph} requires listing the successors of each node on a separate line. If your graph is specified arc by arc (one arc per line) you can use {@link it.unimi.dsi.webgraph.ArcListASCIIGraph} instead. {@link it.unimi.dsi.webgraph.ShiftedByOneArcListASCIIGraph} can be used if your input data numbers (rather insensibly) nodes starting from one.

Importing your labelled data

Arc-labelled graphs are represented using implementations of {@link it.unimi.dsi.webgraph.labelling.ArcLabelledImmutableGraph}. Most arc-labelled graphs are based on an underlying {@link it.unimi.dsi.webgraph.ImmutableGraph}, and the {@link it.unimi.dsi.webgraph.labelling.ArcLabelledImmutableGraph} implementation just provides label handling. The example {@link it.unimi.dsi.webgraph.examples.IntegerTriplesArcLabelledImmutableGraph} shows how to expose your data as an instance of {@link it.unimi.dsi.webgraph.labelling.ArcLabelledImmutableGraph}, so you can save your data using your preferred combination of implementations.

Dependencies

WebGraph requires Java ≥6 and relies on fastutil 6.4 or greater for high-performance containers and algorithms, on the COLT distribution for statistics, on the DSI utilities for bit-level I/O, on Sux4J for succinct data structures, on JSAP for line-command parsing and on log4j for logging.

Note that in principle the DSI utilities depend on a number of additional useful libraries from the Jakarta commons project, including collections, lang, configuration and io.