Release Notes¶

0.12.1¶

Prelude¶

This release is a bugfix release that fixes bugs identified since the 0.12.0 release.

Bug Fixes¶

Fixed a compatibility issue when building rustworkx-core with priority-queue 1.3.0. Fixed #744

Fixed an issue with the custom sequence return types, BFSSuccessors, NodeIndices, EdgeList, WeightedEdgeList, EdgeIndices, and Chains where they previosuly were missing certain attributes that prevented them being used as a sequence for certain built-in functions such as reversed(). Fixed #696.

0.7.1¶

This release includes a fix for an oversight in the previous 0.7.0 and 0.6.0 releases. Those releases both added custom return types BFSSuccessors, NodeIndices, EdgeList, and WeightedEdgeList that implemented the Python sequence protocol which were used in place of lists for certain functions and methods. However, none of those classes had support for being pickled, which was causing compatibility issues for users that were using the return in a context where it would be pickled (for example as an argument to or return of a function called with multiprocessing). This release has a single change over 0.7.0 which is to add the missing support for pickling BFSSuccessors, NodeIndices, EdgeList, and WeightedEdgeList which fixes that issue.

0.7.0¶

This release includes several new features and bug fixes.

This release also dropped support for Python 3.5. If you want to use rustworkx with Python 3.5 that last version which supports Python 3.5 is 0.6.0.

New Features¶

New generator functions for two new generator types, mesh and grid were added to rustworkx.generators for generating all to all and grid graphs respectively. These functions are: mesh_graph(), directed_mesh_graph(), grid_graph(), and directed_grid_graph()
A new function, rustworkx.digraph_union(), for taking the union between two PyDiGraph objects has been added.
A new PyDiGraph method merge_nodes() has been added. This method can be used to merge 2 nodes in a graph if they have the same weight/data payload.
A new PyDiGraph method find_node_by_weight() which can be used to lookup a node index by a given weight/data payload.
A new return type NodeIndices has been added. This class is returned by functions and methods that return a list of node indices. It implements the Python sequence protocol and can be used as list.
Two new return types EdgeList and WeightedEdgeList. These classes are returned from functions and methods that return a list of edge tuples and a list of edge tuples with weights. They both implement the Python sequence protocol and can be used as a list
A new function collect_runs() has been added. This function is used to find linear paths of nodes that match a given condition.

Upgrade Notes¶

Support for running rustworkx on Python 3.5 has been dropped. The last release with support for Python 3.5 is 0.6.0.
The rustworkx.PyDiGraph.node_indexes(), rustworkx.PyDiGraph.neighbors(), rustworkx.PyDiGraph.successor_indices(), rustworkx.PyDiGraph.predecessor_indices(), rustworkx.PyDiGraph.add_nodes_from(), rustworkx.PyGraph.node_indexes(), rustworkx.PyGraph.add_nodes_from(), and rustworkx.PyGraph.neighbors() methods and the dag_longest_path(), topological_sort(), graph_astar_shortest_path(), and digraph_astar_shortest_path() functions now return a NodeIndices object instead of a list of integers. This should not require any changes unless explicit type checking for a list was used.
The rustworkx.PyDiGraph.edge_list(), and rustworkx.PyGraph.edge_list() methods and digraph_dfs_edges(), graph_dfs_edges(), and digraph_find_cycle() functions now return an EdgeList object instead of a list of integers. This should not require any changes unless explicit type checking for a list was used.
The rustworkx.PyDiGraph.weighted_edge_list(), rustworkx.PyDiGraph.in_edges(), rustworkx.PyDiGraph.out_edges(), and rustworkx.PyGraph.weighted_edge_list methods now return a WeightedEdgeList object instead of a list of integers. This should not require any changes unless explicit type checking for a list was used.

Fixes¶

BFSSuccessors objects now can be compared with == and != to any other Python sequence type.
The built and published sdist packages for rustworkx were previously not including the Cargo.lock file. This meant that the reproducible build versions of the rust dependencies were not passed through to source. This has been fixed so building from sdist will always use known working versions that we use for testing in CI.

0.6.0¶

This release includes a number of new features and bug fixes. The main focus of this release was to expand the rustworkx API functionality to include some commonly needed functions that were missing.

This release is also the first release to provide full support for running with Python 3.9. On previous releases Python 3.9 would likely work, but it would require building rustworkx from source. Also this will likely be the final release that supports Python 3.5.

New Features¶

Two new functions, digraph_k_shortest_path_lengths() and graph_k_shortest_path_lengths(), for finding the k shortest path lengths from a node in a PyDiGraph and PyGraph.
A new method, is_symmetric(), to the PyDiGraph class. This method will check whether the graph is symmetric or not
A new kwarg, as_undirected, was added to the digraph_floyd_warshall_numpy() function. This can be used to treat the input PyDiGraph object as if it was undirected for the generated output matrix.
A new function, digraph_find_cycle(), which will return the first cycle during a depth first search of a PyDiGraph object.
Two new functions, directed_gnm_random_graph() and undirected_gnm_random_graph(), for generating random \(G(n, m)\) graphs.
A new method, remove_edges_from(), was added to PyDiGraph and PyGraph (removed_edges_from()). This can be used to remove multiple edges from a graph object in a single call.
A new method, subgraph(), was added to PyDiGraph and PyGraph (subgraph()) which takes in a list of node indices and will return a new object of the same type representing a subgraph containing the node indices in that list.
Support for running with Python 3.9
A new method, to_undirected(), was added to PyDiGraph. This method will generate an undirected PyGraph object from the PyDiGraph object.
A new kwarg, bidirectional, was added to the directed generator functions directed_cycle_graph(), directed_path_graph(), and directed_star_graph(). When set to True the directed graphs generated by these functions will add edges in both directions.
Added two new functions, is_weakly_connected() and weakly_connected_components(), which will either check if a PyDiGraph object is weakly connected or return the list of the weakly connected components of an input PyDiGraph.
The weight_fn kwarg for graph_adjacency_matrix(), digraph_adjacency_matrix(), graph_floyd_warshall_numpy(), and digraph_floyd_warshall_numpy() is now optional. Previously, it always had to be specified when calling these function. But, instead you can now rely on a default weight float (which defaults to 1.0) to be used for all the edges in the graph.
Add a neighbors() method to PyGraph and PyDiGraph (neighbors()). This function will return the node indices of the neighbor nodes for a given input node.
Two new methods, successor_indices() and predecessor_indices(), were added to PyDiGraph. These methods will return the node indices for the successor and predecessor nodes of a given input node.
Two new functions, graph_distance_matrix() and digraph_distance_matrix(), were added for generating a distance matrix from an input PyGraph and PyDiGraph.
Two new functions, digraph_dijkstra_shortest_paths() and graph_dijkstra_shortest_path(), were added for returning the shortest paths from a node in a PyDiGraph and a PyGraph object.
Four new methods, insert_node_on_in_edges(), insert_node_on_out_edges(), insert_node_on_in_edges_multiple(), and insert_node_on_out_edges_multiple() were added to PyDiGraph. These functions are used to insert an existing node in between an reference node(s) and all it’s predecessors or successors.
Two new functions, graph_dfs_edges() and digraph_dfs_edges(), were added to get an edge list in depth first order from a PyGraph and PyDiGraph.

Upgrade Notes¶

The numpy arrays returned by graph_floyd_warshall_numpy(), digraph_floyd_warshall_numpy(), digraph_adjacency_matrix(), and graph_adjacency_matrix() will now be in a contiguous C array memory layout. Previously, they would return arrays in a column-major fortran layout. This was change was made to make it easier to interface the arrays returned by these functions with other C Python extensions. There should be no change when interacting with the numpy arrays via numpy’s API.
The bfs_successors() method now returns an object of a custom type BFSSuccessors instead of a list. The BFSSuccessors type implements the Python sequence protocol so it can be used in place like a list (except for where explicit type checking is used). This was done to defer the type conversion between Rust and Python since doing it all at once can be a performance bottleneck especially for large graphs. The BFSSuccessors class will only do the type conversion when an element is accessed.

Fixes¶

When pickling PyDiGraph objects the original node indices will be preserved across the pickle.
The random \(G(n, p)\) functions, directed_gnp_random_graph() and undirected_gnp_random_graph(), will now also handle exact 0 or 1 probabilities. Previously it would fail in these cases. Fixes #172

0.5.0¶

This release include a number of new features and bug fixes. The main focus of the improvements of this release was to increase the ease of interacting with graph objects. This includes adding support for generating dot output which can be used with graphviz (or similar tools) for visualizing graphs adding more methods to query the state of graph, adding a generator module for easily creating graphs of certain shape, and implementing the mapping protocol so you can directly interact with graph objects.

New Features¶

A new method, to_dot(), was added to PyGraph and PyDiGraph (to_dot()). It will generate a dot format representation of the object which can be used with Graphivz (or similar tooling) to generate visualizations of graphs.
Added a new function, strongly_connected_components(), to get the list of strongly connected components of a PyDiGraph object.
A new method, compose(), for composing another graph object of the same type into a graph was added to PyGraph and PyDiGraph (compose()).
The PyGraph and PyDigraph classes now implement the Python mapping protocol for interacting with graph nodes. You can now access and interact with node data directly by using standard map access patterns in Python. For example, accessing a graph like graph[1] will return the weight/data payload for the node at index 1.
A new module, rustworkx.generators, has been added. Functions in this module can be used for quickly generating graphs of certain shape. To start it includes:
A new method, remove_node_retain_edges(), has been added to the PyDiGraph class. This method can be used to remove a node and add edges from its predecesors to its successors.
Two new methods, edge_list() and weighted_edge_list(), for getting a list of tuples with the edge source and target (with or without edge weights) have been added to PyGraph and PyDiGraph (edge_list() and weighted_edge_list())
A new function, cycle_basis(), for getting a list of cycles which form a basis for cycles of a PyGraph object.
Two new functions, graph_floyd_warshall_numpy() and digraph_floyd_warshall_numpy(), were added for running the Floyd Warshall algorithm and returning all the shortest path lengths as a distance matrix.
A new constructor method, read_edge_list(), has been added to PyGraph and PyDigraph (read_edge_list()). This method will take in a path to an edge list file and will read that file and generate a new object from the contents.
Two new methods, extend_from_edge_list() and extend_from_weighted_edge_list() has been added to PyGraph and PyDiGraph (extend_from_edge_list() and extend_from_weighted_edge_list()). This method takes in an edge list and will add both the edges and nodes (if a node index used doesn’t exist yet) in the list to the graph.

Fixes¶

The limitation with the is_isomorphic() and is_isomorphic_node_match() functions that would cause segfaults when comparing graphs with node removals has been fixed. You can now run either function with any PyDiGraph/PyDAG objects, even if there are node removals. Fixes #27
If an invalid node index was passed as part of the first_layer argument to the layers() function it would previously raise a PanicException that included a Rust backtrace and no other user actionable details which was caused by an unhandled error. This has been fixed so that an IndexError is raised and the problematic node index is included in the exception message.

0.4.0¶

This release includes many new features and fixes, including improved performance and better documentation. But, the biggest change for this release is that this is the first release of rustworkx that supports compilation with a stable released version of rust. This was made possible thanks to all the hard work of the PyO3 maintainers and contributors in the PyO3 0.11.0 release.

New Features¶

A new class for undirected graphs, PyGraph, was added.
2 new functions graph_adjacency_matrix() and digraph_adjacency_matrix() to get the adjacency matrix of a PyGraph and PyDiGraph object.
A new PyDiGraph method, find_adjacent_node_by_edge(), was added. This is used to locate an adjacent node given a condition based on the edge between them.
New methods, add_nodes_from(), add_edges_from(), add_edges_from_no_data(), and remove_nodes_from() were added to PyDiGraph. These methods allow for the addition (and removal) of multiple nodes or edges from a graph in a single call.
A new function, graph_greedy_color(), which is used to return a coloring map from a PyGraph object.
2 new functions, graph_astar_shortest_path() and digraph_astar_shortest_path(), to find the shortest path from a node to a specified goal using the A* search algorithm.
2 new functions, graph_all_simple_paths() and digraph_all_simple_paths(), to return a list of all the simple paths between 2 nodes in a PyGraph or a PyDiGraph object.
2 new functions, directed_gnp_random_graph() and undirected_gnp_random_graph(), to generate \(G_{np}\) random PyDiGraph and PyGraph objects.
2 new functions, graph_dijkstra_shortest_path_lengths() and digraph_dijkstra_shortest_path_lengths(), were added for find the shortest path length between nodes in PyGraph or PyDiGraph object using Dijkstra’s algorithm.

Upgrade Notes¶

The PyDAG class was renamed PyDiGraph to better reflect it’s functionality. For backwards compatibility PyDAG still exists as a Python subclass of PyDiGraph. No changes should be required for existing users.
numpy is now a dependency of rustworkx. This is used for the adjacency matrix functions to return numpy arrays. The minimum version of numpy supported is 1.16.0.

Fixes¶

The rustworkx exception classes are now properly exported from the rustworkx module. In prior releases it was not possible to import the exception classes (normally to catch one being raised) requiring users to catch the base Exception class. This has been fixed so a specialized rustworkx exception class can be used.