mz_compute/render/join/

mz_join_core.rs

// Copyright Materialize, Inc. and contributors. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.

//! A fork of DD's `JoinCore::join_core`.
//!
//! Currently, compute rendering knows two implementations for linear joins:
//!
//!  * Differential's `JoinCore::join_core`
//!  * A Materialize fork thereof, called `mz_join_core`
//!
//! `mz_join_core` exists to solve a responsiveness problem with the DD implementation.
//! DD's join is only able to yield between keys. When computing a large cross-join or a highly
//! skewed join, this can result in loss of interactivity when the join operator refuses to yield
//! control for multiple seconds or longer, which in turn causes degraded user experience.
//!
//! `mz_join_core` currently fixes the yielding issue by omitting the merge-join matching strategy
//! implemented in DD's join implementation. This leaves only the nested loop strategy for which it
//! is easy to implement yielding within keys.
//!
//! While `mz_join_core` retains responsiveness in the face of cross-joins it is also, due to its
//! sole reliance on nested-loop matching, significantly slower than DD's join for workloads that
//! have a large amount of edits at different times. We consider these niche workloads for
//! Materialize today, due to the way source ingestion works, but that might change in the future.
//!
//! For the moment, we keep both implementations around, selectable through a feature flag.
//! We expect `mz_join_core` to be more useful in Materialize today, but being able to fall back to
//! DD's implementation provides a safety net in case that assumption is wrong.
//!
//! In the mid-term, we want to arrive at a single join implementation that is as efficient as DD's
//! join and as responsive as `mz_join_core`. Whether that means adding merge-join matching to
//! `mz_join_core` or adding better fueling to DD's join implementation is still TBD.

use std::cmp::Ordering;
use std::collections::VecDeque;
use std::time::Instant;

use differential_dataflow::Data;
use differential_dataflow::IntoOwned;
use differential_dataflow::consolidation::{consolidate, consolidate_updates};
use differential_dataflow::difference::Multiply;
use differential_dataflow::lattice::Lattice;
use differential_dataflow::operators::arrange::arrangement::Arranged;
use differential_dataflow::trace::{BatchReader, Cursor, TraceReader};
use mz_repr::Diff;
use timely::PartialOrder;
use timely::container::{CapacityContainerBuilder, PushInto, SizableContainer};
use timely::dataflow::channels::pact::Pipeline;
use timely::dataflow::channels::pushers::Tee;
use timely::dataflow::channels::pushers::buffer::Session;
use timely::dataflow::operators::generic::OutputHandleCore;
use timely::dataflow::operators::{Capability, Operator};
use timely::dataflow::{Scope, StreamCore};
use timely::progress::timestamp::Timestamp;
use tracing::trace;

use crate::render::context::ShutdownToken;

/// Joins two arranged collections with the same key type.
///
/// Each matching pair of records `(key, val1)` and `(key, val2)` are subjected to the `result` function,
/// which produces something implementing `IntoIterator`, where the output collection will have an entry for
/// every value returned by the iterator.
pub(super) fn mz_join_core<G, Tr1, Tr2, L, I, YFn, C>(
    arranged1: &Arranged<G, Tr1>,
    arranged2: &Arranged<G, Tr2>,
    shutdown_token: ShutdownToken,
    mut result: L,
    yield_fn: YFn,
) -> StreamCore<G, C>
where
    G: Scope,
    G::Timestamp: Lattice,
    Tr1: TraceReader<Time = G::Timestamp, Diff = Diff> + Clone + 'static,
    Tr2: for<'a> TraceReader<Key<'a> = Tr1::Key<'a>, Time = G::Timestamp, Diff = Diff>
        + Clone
        + 'static,
    L: FnMut(Tr1::Key<'_>, Tr1::Val<'_>, Tr2::Val<'_>) -> I + 'static,
    I: IntoIterator,
    I::Item: Data,
    YFn: Fn(Instant, usize) -> bool + 'static,
    C: SizableContainer + PushInto<(I::Item, G::Timestamp, Diff)> + Data,
{
    let mut trace1 = arranged1.trace.clone();
    let mut trace2 = arranged2.trace.clone();

    arranged1.stream.binary_frontier(
        &arranged2.stream,
        Pipeline,
        Pipeline,
        "Join",
        move |capability, info| {
            let operator_id = info.global_id;

            // Acquire an activator to reschedule the operator when it has unfinished work.
            let activator = arranged1.stream.scope().activator_for(info.address);

            // Our initial invariants are that for each trace, physical compaction is less or equal the trace's upper bound.
            // These invariants ensure that we can reference observed batch frontiers from `_start_upper` onward, as long as
            // we maintain our physical compaction capabilities appropriately. These assertions are tested as we load up the
            // initial work for the two traces, and before the operator is constructed.

            // Acknowledged frontier for each input.
            // These two are used exclusively to track batch boundaries on which we may want/need to call `cursor_through`.
            // They will drive our physical compaction of each trace, and we want to maintain at all times that each is beyond
            // the physical compaction frontier of their corresponding trace.
            // Should we ever *drop* a trace, these are 1. much harder to maintain correctly, but 2. no longer used.
            use timely::progress::frontier::Antichain;
            let mut acknowledged1 = Antichain::from_elem(<G::Timestamp>::minimum());
            let mut acknowledged2 = Antichain::from_elem(<G::Timestamp>::minimum());

            // deferred work of batches from each input.
            let mut todo1 = VecDeque::new();
            let mut todo2 = VecDeque::new();

            // We'll unload the initial batches here, to put ourselves in a less non-deterministic state to start.
            trace1.map_batches(|batch1| {
                trace!(
                    operator_id,
                    input = 1,
                    lower = ?batch1.lower().elements(),
                    upper = ?batch1.upper().elements(),
                    size = batch1.len(),
                    "pre-loading batch",
                );

                acknowledged1.clone_from(batch1.upper());
                // No `todo1` work here, because we haven't accepted anything into `batches2` yet.
                // It is effectively "empty", because we choose to drain `trace1` before `trace2`.
                // Once we start streaming batches in, we will need to respond to new batches from
                // `input1` with logic that would have otherwise been here. Check out the next loop
                // for the structure.
            });
            // At this point, `ack1` should exactly equal `trace1.read_upper()`, as they are both determined by
            // iterating through batches and capturing the upper bound. This is a great moment to assert that
            // `trace1`'s physical compaction frontier is before the frontier of completed times in `trace1`.
            // TODO: in the case that this does not hold, instead "upgrade" the physical compaction frontier.
            assert!(PartialOrder::less_equal(
                &trace1.get_physical_compaction(),
                &acknowledged1.borrow()
            ));

            trace!(
                operator_id,
                input = 1,
                acknowledged1 = ?acknowledged1.elements(),
                "pre-loading finished",
            );

            // We capture batch2 cursors first and establish work second to avoid taking a `RefCell` lock
            // on both traces at the same time, as they could be the same trace and this would panic.
            let mut batch2_cursors = Vec::new();
            trace2.map_batches(|batch2| {
                trace!(
                    operator_id,
                    input = 2,
                    lower = ?batch2.lower().elements(),
                    upper = ?batch2.upper().elements(),
                    size = batch2.len(),
                    "pre-loading batch",
                );

                acknowledged2.clone_from(batch2.upper());
                batch2_cursors.push((batch2.cursor(), batch2.clone()));
            });
            // At this point, `ack2` should exactly equal `trace2.read_upper()`, as they are both determined by
            // iterating through batches and capturing the upper bound. This is a great moment to assert that
            // `trace2`'s physical compaction frontier is before the frontier of completed times in `trace2`.
            // TODO: in the case that this does not hold, instead "upgrade" the physical compaction frontier.
            assert!(PartialOrder::less_equal(
                &trace2.get_physical_compaction(),
                &acknowledged2.borrow()
            ));

            // Load up deferred work using trace2 cursors and batches captured just above.
            for (batch2_cursor, batch2) in batch2_cursors.into_iter() {
                trace!(
                    operator_id,
                    input = 2,
                    acknowledged1 = ?acknowledged1.elements(),
                    "deferring work for batch",
                );

                // It is safe to ask for `ack1` because we have confirmed it to be in advance of `distinguish_since`.
                let (trace1_cursor, trace1_storage) =
                    trace1.cursor_through(acknowledged1.borrow()).unwrap();
                // We could downgrade the capability here, but doing so is a bit complicated mathematically.
                // TODO: downgrade the capability by searching out the one time in `batch2.lower()` and not
                // in `batch2.upper()`. Only necessary for non-empty batches, as empty batches may not have
                // that property.
                todo2.push_back(Deferred::new(
                    trace1_cursor,
                    trace1_storage,
                    batch2_cursor,
                    batch2.clone(),
                    capability.clone(),
                ));
            }

            trace!(
                operator_id,
                input = 2,
                acknowledged2 = ?acknowledged2.elements(),
                "pre-loading finished",
            );

            // Droppable handles to shared trace data structures.
            let mut trace1_option = Some(trace1);
            let mut trace2_option = Some(trace2);

            move |input1, input2, output| {
                // If the dataflow is shutting down, discard all existing and future work.
                if shutdown_token.in_shutdown() {
                    trace!(operator_id, "shutting down");

                    // Discard data at the inputs.
                    input1.for_each(|_cap, _data| ());
                    input2.for_each(|_cap, _data| ());

                    // Discard queued work.
                    todo1 = Default::default();
                    todo2 = Default::default();

                    // Stop holding on to input traces.
                    trace1_option = None;
                    trace2_option = None;

                    return;
                }

                // 1. Consuming input.
                //
                // The join computation repeatedly accepts batches of updates from each of its inputs.
                //
                // For each accepted batch, it prepares a work-item to join the batch against previously "accepted"
                // updates from its other input. It is important to track which updates have been accepted, because
                // we use a shared trace and there may be updates present that are in advance of this accepted bound.
                //
                // Batches are accepted: 1. in bulk at start-up (above), 2. as we observe them in the input stream,
                // and 3. if the trace can confirm a region of empty space directly following our accepted bound.
                // This last case is a consequence of our inability to transmit empty batches, as they may be formed
                // in the absence of timely dataflow capabilities.

                // Drain input 1, prepare work.
                input1.for_each(|capability, data| {
                    let trace2 = trace2_option
                        .as_mut()
                        .expect("we only drop a trace in response to the other input emptying");
                    let capability = capability.retain();
                    for batch1 in data.drain(..) {
                        // Ignore any pre-loaded data.
                        if PartialOrder::less_equal(&acknowledged1, batch1.lower()) {
                            trace!(
                                operator_id,
                                input = 1,
                                lower = ?batch1.lower().elements(),
                                upper = ?batch1.upper().elements(),
                                size = batch1.len(),
                                "loading batch",
                            );

                            if !batch1.is_empty() {
                                trace!(
                                    operator_id,
                                    input = 1,
                                    acknowledged2 = ?acknowledged2.elements(),
                                    "deferring work for batch",
                                );

                                // It is safe to ask for `ack2` as we validated that it was at least `get_physical_compaction()`
                                // at start-up, and have held back physical compaction ever since.
                                let (trace2_cursor, trace2_storage) =
                                    trace2.cursor_through(acknowledged2.borrow()).unwrap();
                                let batch1_cursor = batch1.cursor();
                                todo1.push_back(Deferred::new(
                                    batch1_cursor,
                                    batch1.clone(),
                                    trace2_cursor,
                                    trace2_storage,
                                    capability.clone(),
                                ));
                            }

                            // To update `acknowledged1` we might presume that `batch1.lower` should equal it, but we
                            // may have skipped over empty batches. Still, the batches are in-order, and we should be
                            // able to just assume the most recent `batch1.upper`
                            debug_assert!(PartialOrder::less_equal(&acknowledged1, batch1.upper()));
                            acknowledged1.clone_from(batch1.upper());

                            trace!(
                                operator_id,
                                input = 1,
                                acknowledged1 = ?acknowledged1.elements(),
                                "batch acknowledged",
                            );
                        }
                    }
                });

                // Drain input 2, prepare work.
                input2.for_each(|capability, data| {
                    let trace1 = trace1_option
                        .as_mut()
                        .expect("we only drop a trace in response to the other input emptying");
                    let capability = capability.retain();
                    for batch2 in data.drain(..) {
                        // Ignore any pre-loaded data.
                        if PartialOrder::less_equal(&acknowledged2, batch2.lower()) {
                            trace!(
                                operator_id,
                                input = 2,
                                lower = ?batch2.lower().elements(),
                                upper = ?batch2.upper().elements(),
                                size = batch2.len(),
                                "loading batch",
                            );

                            if !batch2.is_empty() {
                                trace!(
                                    operator_id,
                                    input = 2,
                                    acknowledged1 = ?acknowledged1.elements(),
                                    "deferring work for batch",
                                );

                                // It is safe to ask for `ack1` as we validated that it was at least `get_physical_compaction()`
                                // at start-up, and have held back physical compaction ever since.
                                let (trace1_cursor, trace1_storage) =
                                    trace1.cursor_through(acknowledged1.borrow()).unwrap();
                                let batch2_cursor = batch2.cursor();
                                todo2.push_back(Deferred::new(
                                    trace1_cursor,
                                    trace1_storage,
                                    batch2_cursor,
                                    batch2.clone(),
                                    capability.clone(),
                                ));
                            }

                            // To update `acknowledged2` we might presume that `batch2.lower` should equal it, but we
                            // may have skipped over empty batches. Still, the batches are in-order, and we should be
                            // able to just assume the most recent `batch2.upper`
                            debug_assert!(PartialOrder::less_equal(&acknowledged2, batch2.upper()));
                            acknowledged2.clone_from(batch2.upper());

                            trace!(
                                operator_id,
                                input = 2,
                                acknowledged2 = ?acknowledged2.elements(),
                                "batch acknowledged",
                            );
                        }
                    }
                });

                // Advance acknowledged frontiers through any empty regions that we may not receive as batches.
                if let Some(trace1) = trace1_option.as_mut() {
                    trace!(
                        operator_id,
                        input = 1,
                        acknowledged1 = ?acknowledged1.elements(),
                        "advancing trace upper",
                    );
                    trace1.advance_upper(&mut acknowledged1);
                }
                if let Some(trace2) = trace2_option.as_mut() {
                    trace!(
                        operator_id,
                        input = 2,
                        acknowledged2 = ?acknowledged2.elements(),
                        "advancing trace upper",
                    );
                    trace2.advance_upper(&mut acknowledged2);
                }

                // 2. Join computation.
                //
                // For each of the inputs, we do some amount of work (measured in terms of number
                // of output records produced). This is meant to yield control to allow downstream
                // operators to consume and reduce the output, but it it also means to provide some
                // degree of responsiveness. There is a potential risk here that if we fall behind
                // then the increasing queues hold back physical compaction of the underlying traces
                // which results in unintentionally quadratic processing time (each batch of either
                // input must scan all batches from the other input).

                // Perform some amount of outstanding work.
                trace!(
                    operator_id,
                    input = 1,
                    work_left = todo1.len(),
                    "starting work",
                );

                let start_time = Instant::now();
                let mut work = 0;
                while !todo1.is_empty() && !yield_fn(start_time, work) {
                    todo1.front_mut().unwrap().work(
                        output,
                        &mut result,
                        |w| yield_fn(start_time, w),
                        &mut work,
                    );
                    if !todo1.front().unwrap().work_remains() {
                        todo1.pop_front();
                    }
                }

                trace!(
                    operator_id,
                    input = 1,
                    work_left = todo1.len(),
                    work_done = work,
                    elapsed = ?start_time.elapsed(),
                    "ceasing work",
                );

                // Perform some amount of outstanding work.
                trace!(
                    operator_id,
                    input = 2,
                    work_left = todo2.len(),
                    "starting work",
                );

                let start_time = Instant::now();
                let mut work = 0;
                while !todo2.is_empty() && !yield_fn(start_time, work) {
                    todo2.front_mut().unwrap().work(
                        output,
                        &mut result,
                        |w| yield_fn(start_time, w),
                        &mut work,
                    );
                    if !todo2.front().unwrap().work_remains() {
                        todo2.pop_front();
                    }
                }

                trace!(
                    operator_id,
                    input = 2,
                    work_left = todo2.len(),
                    work_done = work,
                    elapsed = ?start_time.elapsed(),
                    "ceasing work",
                );

                // Re-activate operator if work remains.
                if !todo1.is_empty() || !todo2.is_empty() {
                    activator.activate();
                }

                // 3. Trace maintenance.
                //
                // Importantly, we use `input.frontier()` here rather than `acknowledged` to track
                // the progress of an input, because should we ever drop one of the traces we will
                // lose the ability to extract information from anything other than the input.
                // For example, if we dropped `trace2` we would not be able to use `advance_upper`
                // to keep `acknowledged2` up to date wrt empty batches, and would hold back logical
                // compaction of `trace1`.

                // Maintain `trace1`. Drop if `input2` is empty, or advance based on future needs.
                if let Some(trace1) = trace1_option.as_mut() {
                    if input2.frontier().is_empty() {
                        trace!(operator_id, input = 1, "dropping trace handle");
                        trace1_option = None;
                    } else {
                        trace!(
                            operator_id,
                            input = 1,
                            logical = ?*input2.frontier().frontier(),
                            physical = ?acknowledged1.elements(),
                            "advancing trace compaction",
                        );

                        // Allow `trace1` to compact logically up to the frontier we may yet receive,
                        // in the opposing input (`input2`). All `input2` times will be beyond this
                        // frontier, and joined times only need to be accurate when advanced to it.
                        trace1.set_logical_compaction(input2.frontier().frontier());
                        // Allow `trace1` to compact physically up to the upper bound of batches we
                        // have received in its input (`input1`). We will not require a cursor that
                        // is not beyond this bound.
                        trace1.set_physical_compaction(acknowledged1.borrow());
                    }
                }

                // Maintain `trace2`. Drop if `input1` is empty, or advance based on future needs.
                if let Some(trace2) = trace2_option.as_mut() {
                    if input1.frontier().is_empty() {
                        trace!(operator_id, input = 2, "dropping trace handle");
                        trace2_option = None;
                    } else {
                        trace!(
                            operator_id,
                            input = 2,
                            logical = ?*input1.frontier().frontier(),
                            physical = ?acknowledged2.elements(),
                            "advancing trace compaction",
                        );

                        // Allow `trace2` to compact logically up to the frontier we may yet receive,
                        // in the opposing input (`input1`). All `input1` times will be beyond this
                        // frontier, and joined times only need to be accurate when advanced to it.
                        trace2.set_logical_compaction(input1.frontier().frontier());
                        // Allow `trace2` to compact physically up to the upper bound of batches we
                        // have received in its input (`input2`). We will not require a cursor that
                        // is not beyond this bound.
                        trace2.set_physical_compaction(acknowledged2.borrow());
                    }
                }
            }
        },
    )
}

/// Deferred join computation.
///
/// The structure wraps cursors which allow us to play out join computation at whatever rate we like.
/// This allows us to avoid producing and buffering massive amounts of data, without giving the timely
/// dataflow system a chance to run operators that can consume and aggregate the data.
struct Deferred<T, C1, C2, D>
where
    T: Timestamp,
    C1: Cursor<Time = T, Diff = Diff>,
    C2: for<'a> Cursor<Key<'a> = C1::Key<'a>, Time = T, Diff = Diff>,
{
    cursor1: C1,
    storage1: C1::Storage,
    cursor2: C2,
    storage2: C2::Storage,
    capability: Capability<T>,
    done: bool,
    temp: Vec<(D, T, Diff)>,
}

impl<T, C1, C2, D> Deferred<T, C1, C2, D>
where
    T: Timestamp + Lattice,
    C1: Cursor<Time = T, Diff = Diff>,
    C2: for<'a> Cursor<Key<'a> = C1::Key<'a>, Time = T, Diff = Diff>,
    D: Data,
{
    fn new(
        cursor1: C1,
        storage1: C1::Storage,
        cursor2: C2,
        storage2: C2::Storage,
        capability: Capability<T>,
    ) -> Self {
        Deferred {
            cursor1,
            storage1,
            cursor2,
            storage2,
            capability,
            done: false,
            temp: Vec::new(),
        }
    }

    fn work_remains(&self) -> bool {
        !self.done
    }

    /// Process keys until at least `fuel` output tuples produced, or the work is exhausted.
    fn work<L, I, YFn, C>(
        &mut self,
        output: &mut OutputHandleCore<T, CapacityContainerBuilder<C>, Tee<T, C>>,
        mut logic: L,
        yield_fn: YFn,
        work: &mut usize,
    ) where
        I: IntoIterator<Item = D>,
        L: FnMut(C1::Key<'_>, C1::Val<'_>, C2::Val<'_>) -> I,
        YFn: Fn(usize) -> bool,
        C: SizableContainer + PushInto<(D, T, Diff)> + Data,
    {
        let meet = self.capability.time();

        let mut session = output.session(&self.capability);

        let storage1 = &self.storage1;
        let storage2 = &self.storage2;

        let cursor1 = &mut self.cursor1;
        let cursor2 = &mut self.cursor2;

        let temp = &mut self.temp;

        let flush = |data: &mut Vec<_>, session: &mut Session<_, _, _>| {
            let old_len = data.len();
            // Consolidating here is important when the join closure produces data that
            // consolidates well, for example when projecting columns.
            consolidate_updates(data);
            let recovered = old_len - data.len();
            session.give_iterator(data.drain(..));
            recovered
        };

        assert_eq!(temp.len(), 0);

        let mut buffer = Vec::default();

        while cursor1.key_valid(storage1) && cursor2.key_valid(storage2) {
            match cursor1.key(storage1).cmp(&cursor2.key(storage2)) {
                Ordering::Less => cursor1.seek_key(storage1, cursor2.key(storage2)),
                Ordering::Greater => cursor2.seek_key(storage2, cursor1.key(storage1)),
                Ordering::Equal => {
                    // Populate `temp` with the results, until we should yield.
                    let key = cursor2.key(storage2);
                    while let Some(val1) = cursor1.get_val(storage1) {
                        while let Some(val2) = cursor2.get_val(storage2) {
                            // Evaluate logic on `key, val1, val2`. Note the absence of time and diff.
                            let mut result = logic(key, val1, val2).into_iter().peekable();

                            // We can only produce output if the result return something.
                            if let Some(first) = result.next() {
                                // Join times.
                                cursor1.map_times(storage1, |time1, diff1| {
                                    let mut time1 = time1.into_owned();
                                    time1.join_assign(meet);
                                    let diff1 = diff1.into_owned();
                                    cursor2.map_times(storage2, |time2, diff2| {
                                        let mut time2 = time2.into_owned();
                                        time2.join_assign(&time1);
                                        let diff = diff1.multiply(&diff2.into_owned());
                                        buffer.push((time2, diff));
                                    });
                                });
                                consolidate(&mut buffer);

                                // Special case no results, one result, and potentially many results
                                match (result.peek().is_some(), buffer.len()) {
                                    // Certainly no output
                                    (_, 0) => {}
                                    // Single element, single time
                                    (false, 1) => {
                                        let (time, diff) = buffer.pop().unwrap();
                                        temp.push((first, time, diff));
                                    }
                                    // Multiple elements or multiple times
                                    (_, _) => {
                                        for d in std::iter::once(first).chain(result) {
                                            temp.extend(buffer.iter().map(|(time, diff)| {
                                                (d.clone(), time.clone(), diff.clone())
                                            }))
                                        }
                                    }
                                }
                                buffer.clear();
                            }
                            cursor2.step_val(storage2);
                        }
                        cursor1.step_val(storage1);
                        cursor2.rewind_vals(storage2);

                        *work = work.saturating_add(temp.len());

                        if yield_fn(*work) {
                            // Returning here is only allowed because we leave the cursors in a
                            // state that will let us pick up the work correctly on the next
                            // invocation.
                            *work -= flush(temp, &mut session);
                            if yield_fn(*work) {
                                return;
                            }
                        }
                    }

                    cursor1.step_key(storage1);
                    cursor2.step_key(storage2);
                }
            }
        }

        if !temp.is_empty() {
            *work -= flush(temp, &mut session);
        }

        // We only get here after having iterated through all keys.
        self.done = true;
    }
}