16505110840

Committed 24 Jul 2025 06:40PM UTC coverage: 81.498% (+0.04%) from 81.457%

Build # 16505110840

Build Type

push

github

Committed by

web-flow

Commit Message

Remove scan async API (#4005)

Signed-off-by: Nicholas Gates <nick@nickgates.com>

Run Details

150 of 222 new or added lines in 10 files covered. (67.57%)

10 existing lines in 2 files now uncovered.

42586 of 52254 relevant lines covered (81.5%)

172411.37 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

0.0

/vortex-scan/src/multi_thread.rs

// SPDX-License-Identifier: Apache-2.0
// SPDX-FileCopyrightText: Copyright the Vortex contributors

use std::iter;
use std::sync::{Arc, LazyLock};

use futures::executor::block_on;
use futures::{StreamExt, stream};
use tokio::runtime::{Builder, Runtime};
use vortex_array::ArrayRef;
use vortex_array::iter::{ArrayIterator, ArrayIteratorAdapter};
use vortex_error::{VortexExpect, VortexResult, vortex_err};

use crate::ScanBuilder;

/// We create an internal Tokio runtime used exclusively for orchestrating work-stealing
/// of CPU-bound work for multithreaded scans.
///
/// It is intentionally not exposed to the user, not configurable, and does not enable I/O or
/// timers.
static CPU_RUNTIME: LazyLock<Runtime> = LazyLock::new(|| {
    Builder::new_multi_thread()
        .thread_name("vortex-multithread-scan")
        .build()
        .vortex_expect("Failed to create a new Tokio runtime")
});

impl ScanBuilder<ArrayRef> {
    /// Execute the scan on multiple worker threads.
    pub fn into_array_iter_multithread(self) -> VortexResult<impl ArrayIterator + Send + 'static> {
        let dtype = self.dtype()?;
        Ok(ArrayIteratorAdapter::new(
            dtype,
            self.into_iter_multithread(|a| a)?,
        ))
    }

    /// Execute the scan on multiple worker threads.
    ///
    /// A `map_fn` can be passed to further transform the results of the scan while still running
    /// on the thread pool.
    pub fn into_iter_multithread<T, F>(
        self,
        map_fn: F,
    ) -> VortexResult<impl Iterator<Item = T> + Send + 'static>
    where
        T: 'static + Send,
        F: Fn(VortexResult<ArrayRef>) -> T + Send + Sync + 'static,
    {
        let concurrency = self.concurrency;
        let num_workers = CPU_RUNTIME.metrics().num_workers();

        let tasks = self.build()?;

        // We need to clone and send the map_fn into each task.
        let map_fn = Arc::new(map_fn);

        let handle = CPU_RUNTIME.handle().clone();
        let mut stream = stream::iter(tasks)
            .map(move |task| {
                let map_fn = map_fn.clone();
                // We don't _need_ to spawn the work here. But it allows Tokio to make progress on
                // the tasks in the background, even if the consumer thread is not calling
                // poll_next.
                handle.spawn(async move { task.await.transpose().map(|t| map_fn(t)) })
            })
            // TODO(ngates): this is very crude indeed. This buffered call essentially controls how
            //  many splits we have in-flight at any given time. We multiple workers by concurrency
            //  to configure per-thread concurrency, which essentially means each thread can make
            //  progress on one split while waiting for the I/O of another split to complete.
            //  In an ideal world, the number of in-flight tasks would be dynamically adjusted
            //  based on how much I/O the tasks _actually_ require. For example, all pruning tasks
            //  could be spawned immediately since they all use a single segment, this would allow
            //  head-room to run ahead and figure out the I/O demands of subsequent tasks.
            .buffered(num_workers * concurrency);

        Ok(
            iter::from_fn(move || block_on(stream.next())).filter_map(|result| {
                result
                    .map_err(|e| vortex_err!("Failed to join on a spawned scan task {e}"))
                    .vortex_expect("Failed to join on a spawned scan task")
            }),
        )
    }
}

1	// SPDX-License-Identifier: Apache-2.0
2	// SPDX-FileCopyrightText: Copyright the Vortex contributors
3
4	use std::iter;
5	use std::sync::{Arc, LazyLock};
6
7	use futures::executor::block_on;
8	use futures::{StreamExt, stream};
9	use tokio::runtime::{Builder, Runtime};
10	use vortex_array::ArrayRef;
11	use vortex_array::iter::{ArrayIterator, ArrayIteratorAdapter};
12	use vortex_error::{VortexExpect, VortexResult, vortex_err};
13
14	use crate::ScanBuilder;
15
16	/// We create an internal Tokio runtime used exclusively for orchestrating work-stealing
17	/// of CPU-bound work for multithreaded scans.
18	///
19	/// It is intentionally not exposed to the user, not configurable, and does not enable I/O or
20	/// timers.
NEW 21	static CPU_RUNTIME: LazyLock<Runtime> = LazyLock::new(\|\| {	×
NEW 22	Builder::new_multi_thread()	×
NEW 23	.thread_name("vortex-multithread-scan")	×
NEW 24	.build()	×
NEW 25	.vortex_expect("Failed to create a new Tokio runtime")	×
NEW 26	});	×
27
28	impl ScanBuilder<ArrayRef> {
29	/// Execute the scan on multiple worker threads.
NEW 30	pub fn into_array_iter_multithread(self) -> VortexResult<impl ArrayIterator + Send + 'static> {	×
NEW 31	let dtype = self.dtype()?;	×
NEW 32	Ok(ArrayIteratorAdapter::new(	×
NEW 33	dtype,	×
NEW 34	self.into_iter_multithread(\|a\| a)?,	×
35	))
NEW 36	}	×
37
38	/// Execute the scan on multiple worker threads.
39	///
40	/// A `map_fn` can be passed to further transform the results of the scan while still running
41	/// on the thread pool.
NEW 42	pub fn into_iter_multithread<T, F>(	×
NEW 43	self,	×
NEW 44	map_fn: F,	×
NEW 45	) -> VortexResult<impl Iterator<Item = T> + Send + 'static>	×
NEW 46	where	×
NEW 47	T: 'static + Send,	×
NEW 48	F: Fn(VortexResult<ArrayRef>) -> T + Send + Sync + 'static,	×
49	{
NEW 50	let concurrency = self.concurrency;	×
NEW 51	let num_workers = CPU_RUNTIME.metrics().num_workers();	×
52
NEW 53	let tasks = self.build()?;	×
54
55	// We need to clone and send the map_fn into each task.
NEW 56	let map_fn = Arc::new(map_fn);	×
57
NEW 58	let handle = CPU_RUNTIME.handle().clone();	×
NEW 59	let mut stream = stream::iter(tasks)	×
NEW 60	.map(move \|task\| {	×
NEW 61	let map_fn = map_fn.clone();	×
62	// We don't _need_ to spawn the work here. But it allows Tokio to make progress on
63	// the tasks in the background, even if the consumer thread is not calling
64	// poll_next.
NEW 65	handle.spawn(async move { task.await.transpose().map(\|t\| map_fn(t)) })	×
NEW 66	})	×
67	// TODO(ngates): this is very crude indeed. This buffered call essentially controls how
68	// many splits we have in-flight at any given time. We multiple workers by concurrency
69	// to configure per-thread concurrency, which essentially means each thread can make
70	// progress on one split while waiting for the I/O of another split to complete.
71	// In an ideal world, the number of in-flight tasks would be dynamically adjusted
72	// based on how much I/O the tasks _actually_ require. For example, all pruning tasks
73	// could be spawned immediately since they all use a single segment, this would allow
74	// head-room to run ahead and figure out the I/O demands of subsequent tasks.
NEW 75	.buffered(num_workers * concurrency);	×
76
77	Ok(
NEW 78	iter::from_fn(move \|\| block_on(stream.next())).filter_map(\|result\| {	×
NEW 79	result	×
NEW 80	.map_err(\|e\| vortex_err!("Failed to join on a spawned scan task {e}"))	×
NEW 81	.vortex_expect("Failed to join on a spawned scan task")	×
NEW 82	}),	×
83	)
NEW 84	}	×
85	}

vortex-data / vortex / 16505110840

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous