view rust/hg-core/src/sparse.rs @ 50692:1c31b343e514

match: add `filepath:` pattern to match an exact filepath relative to the root It's useful in certain automated workflows to make sure we recurse in directories whose name conflicts with files in other revisions. In addition it makes it possible to avoid building a potentially costly regex, improving performance when the set of files to match explicitly is large. The benchmark below are run in the following configuration : # data-env-vars.name = mozilla-central-2018-08-01-zstd-sparse-revlog # benchmark.name = files # benchmark.variants.rev = tip # benchmark.variants.files = all-list-filepath-sorted # bin-env-vars.hg.flavor = no-rust It also includes timings using the re2 engine (through the `google-re2` module) to show how much can be saved by just using a better regexp engine. Pattern time (seconds) time using re2 ----------------------------------------------------------- just "." 0.4 0.4 list of "filepath:?" 1.3 1.3 list of "path:?" 25.7 3.9 list of patterns 29.7 10.4 As you can see, Without re2, using "filepath:" instead of "path:" is a huge win. With re2, it is still about three times faster to not have to build the regex.
author Rapha?l Gom?s <rgomes@octobus.net>
date Mon, 12 Jun 2023 16:51:08 +0200
parents e98fd81bb151
children 28c0fcff24e5
line wrap: on
line source

use std::{collections::HashSet, path::Path};

use format_bytes::{write_bytes, DisplayBytes};

use crate::{
    errors::HgError,
    filepatterns::parse_pattern_file_contents,
    matchers::{
        AlwaysMatcher, DifferenceMatcher, IncludeMatcher, Matcher,
        UnionMatcher,
    },
    operations::cat,
    repo::Repo,
    requirements::SPARSE_REQUIREMENT,
    utils::{hg_path::HgPath, SliceExt},
    IgnorePattern, PatternError, PatternFileWarning, PatternSyntax, Revision,
    NULL_REVISION,
};

/// Command which is triggering the config read
#[derive(Copy, Clone, Debug)]
pub enum SparseConfigContext {
    Sparse,
    Narrow,
}

impl DisplayBytes for SparseConfigContext {
    fn display_bytes(
        &self,
        output: &mut dyn std::io::Write,
    ) -> std::io::Result<()> {
        match self {
            SparseConfigContext::Sparse => write_bytes!(output, b"sparse"),
            SparseConfigContext::Narrow => write_bytes!(output, b"narrow"),
        }
    }
}

/// Possible warnings when reading sparse configuration
#[derive(Debug, derive_more::From)]
pub enum SparseWarning {
    /// Warns about improper paths that start with "/"
    RootWarning {
        context: SparseConfigContext,
        line: Vec<u8>,
    },
    /// Warns about a profile missing from the given changelog revision
    ProfileNotFound { profile: Vec<u8>, rev: Revision },
    #[from]
    Pattern(PatternFileWarning),
}

/// Parsed sparse config
#[derive(Debug, Default)]
pub struct SparseConfig {
    // Line-separated
    pub(crate) includes: Vec<u8>,
    // Line-separated
    pub(crate) excludes: Vec<u8>,
    pub(crate) profiles: HashSet<Vec<u8>>,
    pub(crate) warnings: Vec<SparseWarning>,
}

/// All possible errors when reading sparse/narrow config
#[derive(Debug, derive_more::From)]
pub enum SparseConfigError {
    IncludesAfterExcludes {
        context: SparseConfigContext,
    },
    EntryOutsideSection {
        context: SparseConfigContext,
        line: Vec<u8>,
    },
    /// Narrow config does not support '%include' directives
    IncludesInNarrow,
    /// An invalid pattern prefix was given to the narrow spec. Includes the
    /// entire pattern for context.
    InvalidNarrowPrefix(Vec<u8>),
    #[from]
    HgError(HgError),
    #[from]
    PatternError(PatternError),
}

/// Parse sparse config file content.
pub(crate) fn parse_config(
    raw: &[u8],
    context: SparseConfigContext,
) -> Result<SparseConfig, SparseConfigError> {
    let mut includes = vec![];
    let mut excludes = vec![];
    let mut profiles = HashSet::new();
    let mut warnings = vec![];

    #[derive(PartialEq, Eq)]
    enum Current {
        Includes,
        Excludes,
        None,
    }

    let mut current = Current::None;
    let mut in_section = false;

    for line in raw.split(|c| *c == b'\n') {
        let line = line.trim();
        if line.is_empty() || line[0] == b'#' {
            // empty or comment line, skip
            continue;
        }
        if line.starts_with(b"%include ") {
            let profile = line[b"%include ".len()..].trim();
            if !profile.is_empty() {
                profiles.insert(profile.into());
            }
        } else if line == b"[include]" {
            if in_section && current == Current::Includes {
                return Err(SparseConfigError::IncludesAfterExcludes {
                    context,
                });
            }
            in_section = true;
            current = Current::Includes;
            continue;
        } else if line == b"[exclude]" {
            in_section = true;
            current = Current::Excludes;
        } else {
            if current == Current::None {
                return Err(SparseConfigError::EntryOutsideSection {
                    context,
                    line: line.into(),
                });
            }
            if line.trim().starts_with(b"/") {
                warnings.push(SparseWarning::RootWarning {
                    context,
                    line: line.into(),
                });
                continue;
            }
            match current {
                Current::Includes => {
                    includes.push(b'\n');
                    includes.extend(line.iter());
                }
                Current::Excludes => {
                    excludes.push(b'\n');
                    excludes.extend(line.iter());
                }
                Current::None => unreachable!(),
            }
        }
    }

    Ok(SparseConfig {
        includes,
        excludes,
        profiles,
        warnings,
    })
}

fn read_temporary_includes(
    repo: &Repo,
) -> Result<Vec<Vec<u8>>, SparseConfigError> {
    let raw = repo.hg_vfs().try_read("tempsparse")?.unwrap_or_default();
    if raw.is_empty() {
        return Ok(vec![]);
    }
    Ok(raw.split(|c| *c == b'\n').map(ToOwned::to_owned).collect())
}

/// Obtain sparse checkout patterns for the given revision
fn patterns_for_rev(
    repo: &Repo,
    rev: Revision,
) -> Result<Option<SparseConfig>, SparseConfigError> {
    if !repo.has_sparse() {
        return Ok(None);
    }
    let raw = repo.hg_vfs().try_read("sparse")?.unwrap_or_default();

    if raw.is_empty() {
        return Ok(None);
    }

    let mut config = parse_config(&raw, SparseConfigContext::Sparse)?;

    if !config.profiles.is_empty() {
        let mut profiles: Vec<Vec<u8>> = config.profiles.into_iter().collect();
        let mut visited = HashSet::new();

        while let Some(profile) = profiles.pop() {
            if visited.contains(&profile) {
                continue;
            }
            visited.insert(profile.to_owned());

            let output =
                cat(repo, &rev.to_string(), vec![HgPath::new(&profile)])
                    .map_err(|_| {
                        HgError::corrupted(
                            "dirstate points to non-existent parent node"
                                .to_string(),
                        )
                    })?;
            if output.results.is_empty() {
                config.warnings.push(SparseWarning::ProfileNotFound {
                    profile: profile.to_owned(),
                    rev,
                })
            }

            let subconfig = parse_config(
                &output.results[0].1,
                SparseConfigContext::Sparse,
            )?;
            if !subconfig.includes.is_empty() {
                config.includes.push(b'\n');
                config.includes.extend(&subconfig.includes);
            }
            if !subconfig.includes.is_empty() {
                config.includes.push(b'\n');
                config.excludes.extend(&subconfig.excludes);
            }
            config.warnings.extend(subconfig.warnings.into_iter());
            profiles.extend(subconfig.profiles.into_iter());
        }

        config.profiles = visited;
    }

    if !config.includes.is_empty() {
        config.includes.extend(b"\n.hg*");
    }

    Ok(Some(config))
}

/// Obtain a matcher for sparse working directories.
pub fn matcher(
    repo: &Repo,
) -> Result<(Box<dyn Matcher + Sync>, Vec<SparseWarning>), SparseConfigError> {
    let mut warnings = vec![];
    if !repo.requirements().contains(SPARSE_REQUIREMENT) {
        return Ok((Box::new(AlwaysMatcher), warnings));
    }

    let parents = repo.dirstate_parents()?;
    let mut revs = vec![];
    let p1_rev =
        repo.changelog()?
            .rev_from_node(parents.p1.into())
            .map_err(|_| {
                HgError::corrupted(
                    "dirstate points to non-existent parent node".to_string(),
                )
            })?;
    if p1_rev != NULL_REVISION {
        revs.push(p1_rev)
    }
    let p2_rev =
        repo.changelog()?
            .rev_from_node(parents.p2.into())
            .map_err(|_| {
                HgError::corrupted(
                    "dirstate points to non-existent parent node".to_string(),
                )
            })?;
    if p2_rev != NULL_REVISION {
        revs.push(p2_rev)
    }
    let mut matchers = vec![];

    for rev in revs.iter() {
        let config = patterns_for_rev(repo, *rev);
        if let Ok(Some(config)) = config {
            warnings.extend(config.warnings);
            let mut m: Box<dyn Matcher + Sync> = Box::new(AlwaysMatcher);
            if !config.includes.is_empty() {
                let (patterns, subwarnings) = parse_pattern_file_contents(
                    &config.includes,
                    Path::new(""),
                    Some(b"relglob:".as_ref()),
                    false,
                )?;
                warnings.extend(subwarnings.into_iter().map(From::from));
                m = Box::new(IncludeMatcher::new(patterns)?);
            }
            if !config.excludes.is_empty() {
                let (patterns, subwarnings) = parse_pattern_file_contents(
                    &config.excludes,
                    Path::new(""),
                    Some(b"relglob:".as_ref()),
                    false,
                )?;
                warnings.extend(subwarnings.into_iter().map(From::from));
                m = Box::new(DifferenceMatcher::new(
                    m,
                    Box::new(IncludeMatcher::new(patterns)?),
                ));
            }
            matchers.push(m);
        }
    }
    let result: Box<dyn Matcher + Sync> = match matchers.len() {
        0 => Box::new(AlwaysMatcher),
        1 => matchers.pop().expect("1 is equal to 0"),
        _ => Box::new(UnionMatcher::new(matchers)),
    };

    let matcher =
        force_include_matcher(result, &read_temporary_includes(repo)?)?;
    Ok((matcher, warnings))
}

/// Returns a matcher that returns true for any of the forced includes before
/// testing against the actual matcher
fn force_include_matcher(
    result: Box<dyn Matcher + Sync>,
    temp_includes: &[Vec<u8>],
) -> Result<Box<dyn Matcher + Sync>, PatternError> {
    if temp_includes.is_empty() {
        return Ok(result);
    }
    let forced_include_matcher = IncludeMatcher::new(
        temp_includes
            .iter()
            .map(|include| {
                IgnorePattern::new(PatternSyntax::Path, include, Path::new(""))
            })
            .collect(),
    )?;
    Ok(Box::new(UnionMatcher::new(vec![
        Box::new(forced_include_matcher),
        result,
    ])))
}