##// END OF EJS Templates
nodemap: introduce an option to use mmap to read the nodemap mapping...
nodemap: introduce an option to use mmap to read the nodemap mapping The performance and memory benefit is much greater if we don't have to copy all the data in memory for each information. So we introduce an option (on by default) to read the data using mmap. This changeset is the last one definition the API for index support nodemap data. (they have to be able to use the mmaping). Below are some benchmark comparing the best we currently have in 5.3 with the final step of this series (using the persistent nodemap implementation in Rust). The benchmark run `hg perfindex` with various revset and the following variants: Before: * do not use the persistent nodemap * use the CPython implementation of the index for nodemap * use mmapping of the changelog index After: * use the MixedIndex Rust code, with the NodeTree object for nodemap access (still in review) * use the persistent nodemap data from disk * access the persistent nodemap data through mmap * use mmapping of the changelog index The persistent nodemap greatly speed up most operation on very large repositories. Some of the previously very fast lookup end up a bit slower because the persistent nodemap has to be setup. However the absolute slowdown is very small and won't matters in the big picture. Here are some numbers (in seconds) for the reference copy of mozilla-try: Revset Before After abs-change speedup -10000: 0.004622 0.005532 0.000910 × 0.83 -10: 0.000050 0.000132 0.000082 × 0.37 tip 0.000052 0.000085 0.000033 × 0.61 0 + (-10000:) 0.028222 0.005337 -0.022885 × 5.29 0 0.023521 0.000084 -0.023437 × 280.01 (-10000:) + 0 0.235539 0.005308 -0.230231 × 44.37 (-10:) + :9 0.232883 0.000180 -0.232703 ×1293.79 (-10000:) + (:99) 0.238735 0.005358 -0.233377 × 44.55 :99 + (-10000:) 0.317942 0.005593 -0.312349 × 56.84 :9 + (-10:) 0.313372 0.000179 -0.313193 ×1750.68 :9 0.316450 0.000143 -0.316307 ×2212.93 On smaller repositories, the cost of nodemap related operation is not as big, so the win is much more modest. Yet it helps shaving a handful of millisecond here and there. Here are some numbers (in seconds) for the reference copy of mercurial: Revset Before After abs-change speedup -10: 0.000065 0.000097 0.000032 × 0.67 tip 0.000063 0.000078 0.000015 × 0.80 0 0.000561 0.000079 -0.000482 × 7.10 -10000: 0.004609 0.003648 -0.000961 × 1.26 0 + (-10000:) 0.005023 0.003715 -0.001307 × 1.35 (-10:) + :9 0.002187 0.000108 -0.002079 ×20.25 (-10000:) + 0 0.006252 0.003716 -0.002536 × 1.68 (-10000:) + (:99) 0.006367 0.003707 -0.002660 × 1.71 :9 + (-10:) 0.003846 0.000110 -0.003736 ×34.96 :9 0.003854 0.000099 -0.003755 ×38.92 :99 + (-10000:) 0.007644 0.003778 -0.003866 × 2.02 Differential Revision: https://phab.mercurial-scm.org/D7894

File last commit:

r43109:ce6797ef default
r44843:f7459da7 default
Show More
dagops.rs
275 lines | 8.8 KiB | application/rls-services+xml | RustLexer
// dagops.rs
//
// Copyright 2019 Georges Racinet <georges.racinet@octobus.net>
//
// This software may be used and distributed according to the terms of the
// GNU General Public License version 2 or any later version.
//! Miscellaneous DAG operations
//!
//! # Terminology
//! - By *relative heads* of a collection of revision numbers (`Revision`), we
//! mean those revisions that have no children among the collection.
//! - Similarly *relative roots* of a collection of `Revision`, we mean those
//! whose parents, if any, don't belong to the collection.
use super::{Graph, GraphError, Revision, NULL_REVISION};
use crate::ancestors::AncestorsIterator;
use std::collections::{BTreeSet, HashSet};
fn remove_parents(
graph: &impl Graph,
rev: Revision,
set: &mut HashSet<Revision>,
) -> Result<(), GraphError> {
for parent in graph.parents(rev)?.iter() {
if *parent != NULL_REVISION {
set.remove(parent);
}
}
Ok(())
}
/// Relative heads out of some revisions, passed as an iterator.
///
/// These heads are defined as those revisions that have no children
/// among those emitted by the iterator.
///
/// # Performance notes
/// Internally, this clones the iterator, and builds a `HashSet` out of it.
///
/// This function takes an `Iterator` instead of `impl IntoIterator` to
/// guarantee that cloning the iterator doesn't result in cloning the full
/// construct it comes from.
pub fn heads<'a>(
graph: &impl Graph,
iter_revs: impl Clone + Iterator<Item = &'a Revision>,
) -> Result<HashSet<Revision>, GraphError> {
let mut heads: HashSet<Revision> = iter_revs.clone().cloned().collect();
heads.remove(&NULL_REVISION);
for rev in iter_revs {
if *rev != NULL_REVISION {
remove_parents(graph, *rev, &mut heads)?;
}
}
Ok(heads)
}
/// Retain in `revs` only its relative heads.
///
/// This is an in-place operation, so that control of the incoming
/// set is left to the caller.
/// - a direct Python binding would probably need to build its own `HashSet`
/// from an incoming iterable, even if its sole purpose is to extract the
/// heads.
/// - a Rust caller can decide whether cloning beforehand is appropriate
///
/// # Performance notes
/// Internally, this function will store a full copy of `revs` in a `Vec`.
pub fn retain_heads(
graph: &impl Graph,
revs: &mut HashSet<Revision>,
) -> Result<(), GraphError> {
revs.remove(&NULL_REVISION);
// we need to construct an iterable copy of revs to avoid itering while
// mutating
let as_vec: Vec<Revision> = revs.iter().cloned().collect();
for rev in as_vec {
if rev != NULL_REVISION {
remove_parents(graph, rev, revs)?;
}
}
Ok(())
}
/// Roots of `revs`, passed as a `HashSet`
///
/// They are returned in arbitrary order
pub fn roots<G: Graph>(
graph: &G,
revs: &HashSet<Revision>,
) -> Result<Vec<Revision>, GraphError> {
let mut roots: Vec<Revision> = Vec::new();
for rev in revs {
if graph
.parents(*rev)?
.iter()
.filter(|p| **p != NULL_REVISION)
.all(|p| !revs.contains(p))
{
roots.push(*rev);
}
}
Ok(roots)
}
/// Compute the topological range between two collections of revisions
///
/// This is equivalent to the revset `<roots>::<heads>`.
///
/// Currently, the given `Graph` has to implement `Clone`, which means
/// actually cloning just a reference-counted Python pointer if
/// it's passed over through `rust-cpython`. This is due to the internal
/// use of `AncestorsIterator`
///
/// # Algorithmic details
///
/// This is a two-pass swipe inspired from what `reachableroots2` from
/// `mercurial.cext.parsers` does to obtain the same results.
///
/// - first, we climb up the DAG from `heads` in topological order, keeping
/// them in the vector `heads_ancestors` vector, and adding any element of
/// `roots` we find among them to the resulting range.
/// - Then, we iterate on that recorded vector so that a revision is always
/// emitted after its parents and add all revisions whose parents are already
/// in the range to the results.
///
/// # Performance notes
///
/// The main difference with the C implementation is that
/// the latter uses a flat array with bit flags, instead of complex structures
/// like `HashSet`, making it faster in most scenarios. In theory, it's
/// possible that the present implementation could be more memory efficient
/// for very large repositories with many branches.
pub fn range(
graph: &(impl Graph + Clone),
roots: impl IntoIterator<Item = Revision>,
heads: impl IntoIterator<Item = Revision>,
) -> Result<BTreeSet<Revision>, GraphError> {
let mut range = BTreeSet::new();
let roots: HashSet<Revision> = roots.into_iter().collect();
let min_root: Revision = match roots.iter().cloned().min() {
None => {
return Ok(range);
}
Some(r) => r,
};
// Internally, AncestorsIterator currently maintains a `HashSet`
// of all seen revision, which is also what we record, albeit in an ordered
// way. There's room for improvement on this duplication.
let ait = AncestorsIterator::new(graph.clone(), heads, min_root, true)?;
let mut heads_ancestors: Vec<Revision> = Vec::new();
for revres in ait {
let rev = revres?;
if roots.contains(&rev) {
range.insert(rev);
}
heads_ancestors.push(rev);
}
for rev in heads_ancestors.into_iter().rev() {
for parent in graph.parents(rev)?.iter() {
if *parent != NULL_REVISION && range.contains(parent) {
range.insert(rev);
}
}
}
Ok(range)
}
#[cfg(test)]
mod tests {
use super::*;
use crate::testing::SampleGraph;
/// Apply `retain_heads()` to the given slice and return as a sorted `Vec`
fn retain_heads_sorted(
graph: &impl Graph,
revs: &[Revision],
) -> Result<Vec<Revision>, GraphError> {
let mut revs: HashSet<Revision> = revs.iter().cloned().collect();
retain_heads(graph, &mut revs)?;
let mut as_vec: Vec<Revision> = revs.iter().cloned().collect();
as_vec.sort();
Ok(as_vec)
}
#[test]
fn test_retain_heads() -> Result<(), GraphError> {
assert_eq!(retain_heads_sorted(&SampleGraph, &[4, 5, 6])?, vec![5, 6]);
assert_eq!(
retain_heads_sorted(&SampleGraph, &[4, 1, 6, 12, 0])?,
vec![1, 6, 12]
);
assert_eq!(
retain_heads_sorted(&SampleGraph, &[1, 2, 3, 4, 5, 6, 7, 8, 9])?,
vec![3, 5, 8, 9]
);
Ok(())
}
/// Apply `heads()` to the given slice and return as a sorted `Vec`
fn heads_sorted(
graph: &impl Graph,
revs: &[Revision],
) -> Result<Vec<Revision>, GraphError> {
let heads = heads(graph, revs.iter())?;
let mut as_vec: Vec<Revision> = heads.iter().cloned().collect();
as_vec.sort();
Ok(as_vec)
}
#[test]
fn test_heads() -> Result<(), GraphError> {
assert_eq!(heads_sorted(&SampleGraph, &[4, 5, 6])?, vec![5, 6]);
assert_eq!(
heads_sorted(&SampleGraph, &[4, 1, 6, 12, 0])?,
vec![1, 6, 12]
);
assert_eq!(
heads_sorted(&SampleGraph, &[1, 2, 3, 4, 5, 6, 7, 8, 9])?,
vec![3, 5, 8, 9]
);
Ok(())
}
/// Apply `roots()` and sort the result for easier comparison
fn roots_sorted(
graph: &impl Graph,
revs: &[Revision],
) -> Result<Vec<Revision>, GraphError> {
let mut as_vec = roots(graph, &revs.iter().cloned().collect())?;
as_vec.sort();
Ok(as_vec)
}
#[test]
fn test_roots() -> Result<(), GraphError> {
assert_eq!(roots_sorted(&SampleGraph, &[4, 5, 6])?, vec![4]);
assert_eq!(
roots_sorted(&SampleGraph, &[4, 1, 6, 12, 0])?,
vec![0, 4, 12]
);
assert_eq!(
roots_sorted(&SampleGraph, &[1, 2, 3, 4, 5, 6, 7, 8, 9])?,
vec![1, 8]
);
Ok(())
}
/// Apply `range()` and convert the result into a Vec for easier comparison
fn range_vec(
graph: impl Graph + Clone,
roots: &[Revision],
heads: &[Revision],
) -> Result<Vec<Revision>, GraphError> {
range(&graph, roots.iter().cloned(), heads.iter().cloned())
.map(|bs| bs.into_iter().collect())
}
#[test]
fn test_range() -> Result<(), GraphError> {
assert_eq!(range_vec(SampleGraph, &[0], &[4])?, vec![0, 1, 2, 4]);
assert_eq!(range_vec(SampleGraph, &[0], &[8])?, vec![]);
assert_eq!(
range_vec(SampleGraph, &[5, 6], &[10, 11, 13])?,
vec![5, 10]
);
assert_eq!(
range_vec(SampleGraph, &[5, 6], &[10, 12])?,
vec![5, 6, 9, 10, 12]
);
Ok(())
}
}