History

The Boost Graph Library began its life as the Generic Graph Component Library (GGCL), a software project at the Lab for Scientific Computing (LSC) at the University of Notre Dame, under the direction of Professor Andrew Lumsdaine. The Lab’s research directions include numerical linear algebra, parallel computing, and software engineering (including generic programming).

Soon after the Standard Template Library was released, work began at the LSC to apply generic programming to scientific computing. The Matrix Template Library (Jeremy Siek’s masters thesis) was one of the first projects. Many of the lessons learned during construction of the MTL were applied to the design and implementation of the GGCL.

Graph algorithms play an important role in sparse matrix computations, so the LSC had a need for a good graph library. However, none of the available graph libraries (LEDA, GTL, Stanford GraphBase) were written using the generic programming style of the STL, and hence did not fulfill the flexibility and high-performance requirements of the LSC. Others were also expressing interest in a generic C++ graph library. During a meeting with Bjarne Stroustrup we were introduced to several people at AT&T who needed such a library. There had also been earlier work in the area of generic graph algorithms, including some codes written by Alexander Stepanov, and Dietmar Kühl’s masters thesis.

With this in mind, and motivated by homework assignments in his algorithms class, Jeremy began prototyping an interface and some graph classes in the spring of 1998. Lie-Quan Lee then developed the first version of GGCL, which became his masters thesis project.

The following year, Jeremy went to work for SGI with Alexander Stepanov and Matt Austern. During this time Alex’s disjoint-sets based connected components algorithm was added to GGCL, and Jeremy began working on the concept documentation for GGCL similar to Matt’s STL documentation.

While working at SGI, Jeremy heard about Boost and was excited to find a group of people interested in creating high-quality C++ libraries. At Boost there were several people interested in generic graph algorithms, most notably Dietmar Kühl. Some discussions about generic interfaces for graph structures resulted in a revision of GGCL which closely resembles the current Boost Graph Library interface.

On September 4, 2000 GGCL passed the Boost formal review and became the Boost Graph Library (BGL). The first release of BGL was September 27, 2000.

Historical changelog (through Boost 1.36)

This changelog is frozen at Boost 1.36 (2008). Modern release notes live in the Boost release announcements and the git history of boostorg/graph. It is retained here as a record of the library’s early evolution.

Version 1.36.0

New algorithms and components

r_c_shortest_paths, resource-constrained shortest paths, from Michael Drexl.

Version 1.35.0

New algorithms and components

boykov_kolmogorov_max_flow (formerly kolmogorov_max_flow), from Stephan Diederich as part of the 2006 Google Summer of Code.
read_dimacs_max_flow and write_dimacs_max_flow for max-flow problems, from Stephan Diederich.
read_graphml / write_graphml for GraphML input/output, from Tiago de Paula Peixoto.
minimum_cycle_ratio and maximum_cycle_ratio, from Dmitry Bufistov and Andrey Parfenov.
boyer_myrvold_planarity_test, along with a suite of algorithms for planar graphs, from Aaron Windsor.

Enhancements

LEDA adaptor improvements, from Jens Müller.

Version 1.34.1

Bug fixes

bellman_ford_shortest_paths: fixed a bug where certain negative cycles were not correctly detected.

Version 1.34.0

New algorithms and components

edmonds_maximum_cardinality_matching, from Aaron Windsor.
lengauer_tarjan_dominator_tree, from JongSoo Park.
compressed_sparse_row_graph, from Jeremiah Willcock and Douglas Gregor of Indiana University.
sorted_erdos_renyi_iterator, from Jeremiah Willcock of Indiana University.

Enhancements

The compiled library for GraphViz reading is now called boost_graph rather than bgl-viz.
biconnected_components gained a visitor parameter and supports named parameters, from Janusz Piwowarski.
adjacency_matrix now models Bidirectional Graph.
adjacency_list is now Serializable.
Added edges_size_type and vertices_size_type to adjacency_list_traits.
Added operator< etc. to `adjacency_list’s edge descriptor.

Bug fixes

Fixed a bug that caused the relaxed heap to fail on x86 Linux.
Bundled properties now work with adjacency list I/O.
floyd_warshall_all_pairs_shortest_paths now properly uses its compare, inf, and zero parameters.
johnson_all_pairs_shortest_paths now supports compare, combine, inf, and zero.
Fixed a bug in smallest_last_vertex_ordering.hpp that could cause a vertex to be moved to the wrong bucket during a bucket-sorter update.

Version 1.33.1

Bug fixes

fruchterman_reingold_force_directed_layout: fixed enumeration of grid-force pairs, which caused some odd graph formations along grid lines.
king_ordering and cuthill_mckee_ordering: fixed a bug that caused failures with the multi-component version of these algorithms.

Version 1.33.0

New algorithms and components

Experimental Python bindings, from Doug Gregor and Indiana University.
floyd_warshall_all_pairs_shortest_paths, from Lauren Foutz and Scott Hill.
astar_search, from Kristopher Beevers and Jufeng Peng.
fruchterman_reingold_force_directed_layout, from Doug Gregor and Indiana University.
biconnected_components and articulation_points, from Indiana University.
gursoy_atun_layout, from Jeremiah Willcock and Doug Gregor of Indiana University.
king_ordering, from D. Kevin McGrath of Indiana University.
erdos_renyi_iterator.
plod_iterator.
small_world_iterator.

Enhancements

bellman_ford_shortest_paths now permits one to specify the starting vertex, so that it will perform its own initialization.
undirected_dfs is now data-recursive, resulting in improved performance in some cases, from Synge Todo.
dijkstra_shortest_paths now uses a relaxed heap [55] as its priority queue, improving its complexity to O(V log V) and improving real-world performance for larger graphs.
read_graphviz now has a new, Spirit-based parser that works for all graph types and supports arbitrary properties on the graph, from Ron Garcia. The old, Bison-based GraphViz reader has been deprecated and was later removed.
write_graphviz now supports output of dynamic properties (as read in through the new read_graphviz).
cuthill_mckee_ordering has been recast as an invocation of breadth_first_search and now supports graphs with multiple components.
subgraph now supports bundled properties. get_property now refers to the subgraph property, not the root graph’s property.
filtered_graph now supports bundled properties.
reverse_graph now supports bundled properties, set_property, and get_property.

Bug fixes

bellman_ford_shortest_paths now deals with unreachable vertices better.
adjacency_list: parallel edge removal with OutEdgeListS = listS has been fixed. Copying and swapping has been fixed.
Incremental connected components: fixed a bug in the incremental_components routine that may have caused incorrect results.
The remove_out_edge_if function for an undirected adjacency_list has been rewritten and should no longer dereference singular iterators.
write_graphviz now accepts a vertex_id parameter that is used to name the nodes.
read_graphviz now accepts empty attribute lists.
sequential_vertex_coloring has been updated, tested, and documented.

Edit this Page