nxbench.backends package

Submodules

nxbench.backends.core module

class nxbench.backends.core.BackendManager

Bases: object

A registry for dynamically registering and configuring networkx backends.

The registry maps:

backend_name (str) ->

(import_name, conversion_func, teardown_func)

where:

• import_name (str) is the actual module name to import

• conversion_func(nx_graph, num_threads) -> typed backend graph

• teardown_func() -> None (optional)

__init__()

configure_backend(name, original_graph, num_threads)

Convert the given networkx.Graph to the backend-specific representation.

Return type:

Any

Parameters:

• name (str)

• original_graph (Any)

• num_threads (int)

get_version(name)

Retrieve the version for the given backend by its import_name. If not installed, returns “unknown”.

Return type:

str

Parameters:

name (str)

is_available(name)

Check if the registered backend is actually installed (importable).

Return type:

bool

Parameters:

name (str)

is_registered(name)

Return type:

bool

Parameters:

name (str)

register_backend(name, import_name, conversion_func, teardown_func=None)

Register a new backend with a given name, import_name, conversion function, and optional teardown function.

Example:

register_backend(

name=”networkx”, import_name=”networkx”, conversion_func=convert_networkx, teardown_func=teardown_networkx

)

Parameters:

• name (str)

• import_name (str)

• conversion_func (Callable[[Any, int], Any])

• teardown_func (Callable[[], None] | None)

teardown_backend(name)

If a teardown function is registered, call it. Otherwise, do nothing.

Parameters:

name (str)

nxbench.backends.core.get_backend_version(backend_name)

Attempt to retrieve backend_name’s version. - First, check __version__ attribute. - If that fails, fallback to importlib.metadata.get_version(). - If everything fails, return “unknown”.

Return type:

str

Parameters:

backend_name (str)

nxbench.backends.core.is_available(backend_name)

Return True if backend_name is importable; False otherwise. E.g., ‘networkx’, ‘nx_cugraph’, ‘graphblas_algorithms’, etc.

Return type:

bool

Parameters:

backend_name (str)

nxbench.backends.registry module

nxbench.backends.registry.convert_cugraph(original_graph, num_threads)

Parameters:

• original_graph (Graph)

• num_threads (int)

nxbench.backends.registry.convert_graphblas(original_graph, num_threads)

Parameters:

• original_graph (Graph)

• num_threads (int)

nxbench.backends.registry.convert_networkx(original_graph, num_threads)

Parameters:

• original_graph (Graph)

• num_threads (int)

nxbench.backends.registry.convert_parallel(original_graph, num_threads)

Parameters:

• original_graph (Graph)

• num_threads (int)

nxbench.backends.registry.teardown_cugraph()

nxbench.backends.registry.teardown_graphblas()

nxbench.backends.registry.teardown_networkx()

nxbench.backends.registry.teardown_parallel()

Module contents