Struct graph_csr::compute::ComputeGraph

pub struct ComputeGraph<'a, T, DataType> { /* private fields */ }

This is the compute abstraction over a graph. It contains an underlying representation of the data that can support running algorithms over it. Each node contains DataType data, and a status indicating whether or not it is active in the next iteration. Its methods are parallelized using atomics, and will yield good multi-threaded performance.

Implementations

impl<'a, T, DataType> ComputeGraph<'a, T, DataType> where
    T: ValidGraphType + Send + Sync,
    DataType: GraphData, 

pub fn new(graph: &'a Graph<'a, T>) -> Self

Creates a new graph that can run algorithms and can keep track of node data as well as the active nodes.

pub fn set_active(&mut self, idx: usize, status: bool)

Set a single node’s activity in the next iteration as status.

pub fn set_data(&mut self, idx: usize, data: DataType)

Sets a single node’s data in the next iteration as data.

pub fn fill_active(&mut self, status: bool)

Sets all nodes’ activity in the next iteration as status.

pub fn fill_data(&mut self, data: DataType)

Sets all nodes’ data in the next iteration as data.

pub fn step(&mut self)

Performs a global iteration step, useful in many algorithms. The previous status of all nodes is now updated to the new status. The new status is reset to false.

pub fn n_active(&self) -> usize

Returns how many nodes are active in the last iteration. This function calculates the value every time, so it is recommended to store its value.

pub fn push<F>(&mut self, func: F) where
    F: Fn(DataType, &Atomic<DataType>) -> bool + Sync, 

This function iterates over the active nodes in the last iteration and applies func on them. func receives two arguments: old, which is the last state of the node, and new, which is the current state.

pub fn get_data_as_slice(&self) -> &[Atomic<DataType>]

pub fn save_data_to_file(&self, filename: &str) -> Result<()>

Saves the computation’s data to the specified file in binary format, following the local machine’s endianness.