1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
//! General purpose intra-timestamp aggregation
use std::hash::Hash;
use std::collections::HashMap;
use crate::{Data, ExchangeData};
use crate::dataflow::{Stream, Scope};
use crate::dataflow::operators::generic::operator::Operator;
use crate::dataflow::channels::pact::Exchange;
/// Generic intra-timestamp aggregation
///
/// Extension method supporting aggregation of keyed data within timestamp.
/// For inter-timestamp aggregation, consider `StateMachine`.
pub trait Aggregate<S: Scope, K: ExchangeData+Hash, V: ExchangeData> {
/// Aggregates data of the form `(key, val)`, using user-supplied logic.
///
/// The `aggregate` method is implemented for streams of `(K, V)` data,
/// and takes functions `fold`, `emit`, and `hash`; used to combine new `V`
/// data with existing `D` state, to produce `R` output from `D` state, and
/// to route `K` keys, respectively.
///
/// Aggregation happens within each time, and results are produced once the
/// time is complete.
///
/// # Examples
/// ```
/// use timely::dataflow::operators::{ToStream, Map, Inspect};
/// use timely::dataflow::operators::aggregation::Aggregate;
///
/// timely::example(|scope| {
///
/// (0..10).to_stream(scope)
/// .map(|x| (x % 2, x))
/// .aggregate(
/// |_key, val, agg| { *agg += val; },
/// |key, agg: i32| (key, agg),
/// |key| *key as u64
/// )
/// .inspect(|x| assert!(*x == (0, 20) || *x == (1, 25)));
/// });
/// ```
///
/// By changing the type of the aggregate value, one can accumulate into different types.
/// Here we accumulate the data into a `Vec<i32>` and report its length (which we could
/// obviously do more efficiently; imagine we were doing a hash instead).
///
/// ```
/// use timely::dataflow::operators::{ToStream, Map, Inspect};
/// use timely::dataflow::operators::aggregation::Aggregate;
///
/// timely::example(|scope| {
///
/// (0..10).to_stream(scope)
/// .map(|x| (x % 2, x))
/// .aggregate::<_,Vec<i32>,_,_,_>(
/// |_key, val, agg| { agg.push(val); },
/// |key, agg| (key, agg.len()),
/// |key| *key as u64
/// )
/// .inspect(|x| assert!(*x == (0, 5) || *x == (1, 5)));
/// });
/// ```
fn aggregate<R: Data, D: Default+'static, F: Fn(&K, V, &mut D)+'static, E: Fn(K, D)->R+'static, H: Fn(&K)->u64+'static>(
&self,
fold: F,
emit: E,
hash: H) -> Stream<S, R> where S::Timestamp: Eq;
}
impl<S: Scope, K: ExchangeData+Hash+Eq, V: ExchangeData> Aggregate<S, K, V> for Stream<S, (K, V)> {
fn aggregate<R: Data, D: Default+'static, F: Fn(&K, V, &mut D)+'static, E: Fn(K, D)->R+'static, H: Fn(&K)->u64+'static>(
&self,
fold: F,
emit: E,
hash: H) -> Stream<S, R> where S::Timestamp: Eq {
let mut aggregates = HashMap::new();
self.unary_notify(Exchange::new(move |&(ref k, _)| hash(k)), "Aggregate", vec![], move |input, output, notificator| {
// read each input, fold into aggregates
input.for_each(|time, data| {
let agg_time = aggregates.entry(time.time().clone()).or_insert_with(HashMap::new);
for (key, val) in data.drain(..) {
let agg = agg_time.entry(key.clone()).or_insert_with(Default::default);
fold(&key, val, agg);
}
notificator.notify_at(time.retain());
});
// pop completed aggregates, send along whatever
notificator.for_each(|time,_,_| {
if let Some(aggs) = aggregates.remove(time.time()) {
let mut session = output.session(&time);
for (key, agg) in aggs {
session.give(emit(key, agg));
}
}
});
})
}
}