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 on 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 the 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.
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.
Note: the name of the compiled library for GraphViz reading is now called
boost_graph rather than
bgl-viz.
biconnected_components now has a visitor parameter and supports named parameters, from Janusz Piwowarski.
adjacency_matrix now models the Bidirectional Graph concept.
adjacency_list is now Serializable, from Jeremy Siek of Rice University.
Added edges_size_type and vertices_size_type to adjacency_list_traits, from Jeremy Siek of Rice University.
Added operator< , etc. to the edge descriptor of adjacency_list, from Jeremy Siek of Rice University.
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 which could cause a vertex to be moved to the wrong bucket during an BucketSorter update.
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 bug that caused failures with the multi-component version of these algorithms.
Version 1.33.0
New algorithms and components
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.
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 [61] 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 will be removed in a future Boost release.
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 bundlecon_graph_d properties, set_property, and get_property.
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.
| Copyright © 2000-2001 Jeremy Siek, Indiana University ([email protected]) Lie-Quan Lee, Indiana University ([email protected]) Andrew Lumsdaine, Indiana University ([email protected]) |