palimpsest-dataflow 0.1.1

//! Sequential (non-concurrent) graph algorithms.

use std::hash::Hash;

use timely::dataflow::*;

use crate::hashable::Hashable;
use crate::lattice::Lattice;
use crate::operators::*;
use crate::{ExchangeData, VecCollection};

fn _color<G, N>(edges: &VecCollection<G, (N, N)>) -> VecCollection<G, (N, Option<u32>)>
where
    G: Scope<Timestamp: Lattice + Ord + Hash>,
    N: ExchangeData + Hash,
{
    // need some bogus initial values.
    let start = edges.map(|(x, _y)| (x, u32::max_value())).distinct();

    // repeatedly apply color-picking logic.
    sequence(&start, edges, |_node, vals| {
        // look for the first absent positive integer.
        // start at 1 in case we ever use NonZero<u32>.

        (1u32..)
            .find(|&i| vals.get(i as usize - 1).map(|x| *x.0) != Some(i))
            .unwrap()
    })
}

/// Applies `logic` to nodes sequentially, in order of node identifiers.
///
/// The `logic` function updates a node's state as a function of its
/// neighbor states. It will only be called on complete input.
///
/// Internally, this method performs a fixed-point computation in which
/// a node "fires" once all of its neighbors with lower identifier have
/// fired, and we apply `logic` to the new state of lower neighbors and
/// the old state (input) of higher neighbors.
pub fn sequence<G, N, V, F>(
    state: &VecCollection<G, (N, V)>,
    edges: &VecCollection<G, (N, N)>,
    logic: F,
) -> VecCollection<G, (N, Option<V>)>
where
    G: Scope<Timestamp: Lattice + Hash + Ord>,
    N: ExchangeData + Hashable,
    V: ExchangeData,
    F: Fn(&N, &[(&V, isize)]) -> V + 'static,
{
    // start iteration with None messages for all.
    state
        .map(|(node, _state)| (node, None))
        .iterate(|new_state| {
            // immutable content: edges and initial state.
            let edges = edges.enter(&new_state.scope());
            let old_state = state.enter(&new_state.scope());
            // .map(|x| (x.0, Some(x.1)));

            // break edges into forward and reverse directions.
            let forward = edges.filter(|edge| edge.0 < edge.1);
            let reverse = edges.filter(|edge| edge.0 > edge.1);

            // new state goes along forward edges, old state along reverse edges
            let new_messages = new_state.join_map(&forward, |_k, v, d| (d.clone(), v.clone()));

            let incomplete = new_messages
                .filter(|x| x.1.is_none())
                .map(|x| x.0)
                .distinct();
            let new_messages = new_messages
                .filter(|x| x.1.is_some())
                .map(|x| (x.0, x.1.unwrap()));

            let old_messages = old_state.join_map(&reverse, |_k, v, d| (d.clone(), v.clone()));

            let messages = new_messages.concat(&old_messages).antijoin(&incomplete);

            // // determine who has incoming `None` messages, and suppress all of them.
            // let incomplete = new_messages.filter(|x| x.1.is_none()).map(|x| x.0).distinct();

            // merge messages; suppress computation if not all inputs available yet.
            messages
                // .concat(&old_messages)  // /-- possibly too clever: None if any inputs None.
                // .antijoin(&incomplete)
                .reduce(move |k, vs, t| t.push((Some(logic(k,vs)),1)))
                .concat(&incomplete.map(|x| (x, None)))
        })
}