Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[perf] rg does not use globs to prune recursion when it can #2789

Open
1 task done
BGR360 opened this issue Apr 25, 2024 · 4 comments
Open
1 task done

[perf] rg does not use globs to prune recursion when it can #2789

BGR360 opened this issue Apr 25, 2024 · 4 comments
Labels
enhancement An enhancement to the functionality of the software.

Comments

@BGR360
Copy link

BGR360 commented Apr 25, 2024

Please tick this box to confirm you have reviewed the above.

  • I have a different issue.

What version of ripgrep are you using?

ripgrep 14.1.0

How did you install ripgrep?

Build from source with the following patch applied:

diff --git a/crates/core/flags/hiargs.rs b/crates/core/flags/hiargs.rs
index e027a2c..140f006 100644
--- a/crates/core/flags/hiargs.rs
+++ b/crates/core/flags/hiargs.rs
@@ -896,6 +896,7 @@ impl HiArgs {
             .ignore_case_insensitive(self.ignore_file_case_insensitive);
         if !self.no_ignore_dot {
             builder.add_custom_ignore_filename(".rgignore");
+            builder.add_custom_ignore_filename(".hgignore");
         }
         // When we want to sort paths lexicographically in ascending order,
         // then we can actually do this during directory traversal itself.

What operating system are you using ripgrep on?

Linux 5.15.0-60-generic #66~20.04.1-Ubuntu SMP x86_64 GNU/Linux

Describe your bug.

I'm trying to search through a massive corpus of log files (~10M files), on a remote NFS mount, to see if a particular string is present in a certain type of log file. I have a glob that filters down to the log files I care about. The key point is that the files that match my glob are a small subset of all the files.

The corpus looks like this:

Expand for preview of corpus
.
├── Customer1
│   ├── cluster1
│   │   ├── 2024-01-01
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-02-02
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-03-03
│   │   │   └── ...lots-of-logs
│   │   └── ...many-more-dates
│   ├── cluster2
│   │   ├── 2024-01-01
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-02-02
│   │   │   ├── ...lots-of-logs
│   │   │   └── perf_123
│   │   │       └── profile
│   │   │           └── 0-trigger
│   │   │               └── oplogs
│   │   │                   └── SOME-INTERESTING-LOGS
│   │   ├── 2024-03-03
│   │   │   └── ...lots-of-logs
│   │   └── ...many-more-dates
│   └── cluster3
│       ├── 2024-01-01
│       │   └── ...lots-of-logs
│       ├── 2024-02-02
│       │   └── ...lots-of-logs
│       ├── 2024-03-03
│       │   └── ...lots-of-logs
│       └── ...many-more-dates
├── Customer2
│   ├── cluster1
│   │   ├── 2024-01-01
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-02-02
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-03-03
│   │   │   ├── ...lots-of-logs
│   │   │   └── perf_456
│   │   │       └── profile
│   │   │           └── 0-trigger
│   │   │               └── oplogs
│   │   │                   └── SOME-INTERESTING-LOGS
│   │   └── ...many-more-dates
│   ├── cluster2
│   │   ├── 2024-01-01
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-02-02
│   │   │   └── ...lots-of-logs
│   │   ├── 2024-03-03
│   │   │   └── ...lots-of-logs
│   │   └── ...many-more-dates
│   └── cluster3
│       ├── 2024-01-01
│       │   └── ...lots-of-logs
│       ├── 2024-02-02
│       │   └── ...lots-of-logs
│       ├── 2024-03-03
│       │   └── ...lots-of-logs
│       └── ...many-more-dates
├── ...a-few-hundred-more

With my glob being */*/*/perf_*/profile/0-trigger/oplogs/*.log.

The problem is that ripgrep is not limiting its recursive walk to only the paths that definitely match the glob. It is enumerating directories that could not possibly match the glob, and the number of files that end up being considered really adds up. It's considering far more files than it needs to.

What are the steps to reproduce the behavior?

Create the following directory tree. It mimics my corpus.

$ tree
.
├── 0-many
│   ├── blah
│   │   ├── 0-lots
│   │   ├── 1-of
│   │   └── 2-files
│   └── GOOD
│       └── COOL.log
└── 1-dirs
    ├── blah
    │   ├── 0-lots
    │   ├── 1-of
    │   └── 2-files
    └── GOOD
        └── COOL.log

Use the following glob search. It mimics my search. I only want to search through the directories I know will contain my interesting files.

$ rg --debug --files -g '*/GOOD/*.log'

What is the actual behavior?

ripgrep recurses into the */blah/ directories when there's no chance that they could match the glob.

Problematic lines emphasized with >>>

$ rg --debug --files -g '*/GOOD/*.log'
rg: DEBUG|rg::flags::parse|crates/core/flags/parse.rs:97: no extra arguments found from configuration file
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1100: using heuristics to determine whether to read from stdin or search ./ (is_readable_stdin=false, stdin_consumed=false, mode=Files)
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1110: heuristic chose to search ./
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1261: found hostname for hyperlink configuration: timmy
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1271: hyperlink format: ""
rg: DEBUG|globset|crates/globset/src/lib.rs:453: built glob set; 0 literals, 0 basenames, 0 extensions, 0 prefixes, 0 suffixes, 1 required extensions, 0 regexes
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:174: using 12 thread(s)
rg: DEBUG|globset|crates/globset/src/lib.rs:453: built glob set; 0 literals, 3 basenames, 0 extensions, 0 prefixes, 0 suffixes, 0 required extensions, 0 regexes
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./1-dirs/blah/2-files: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./1-dirs/blah/1-of: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./1-dirs/blah/0-lots: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1802: whitelisting ./1-dirs/GOOD/COOL.log: Whitelist(IgnoreMatch(Override(Glob(Matched(Glob { from: None, original: "*/GOOD/*.log", actual: "*/GOOD/*.log", is_whitelist: false, is_only_dir: false })))))
1-dirs/GOOD/COOL.log
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./0-many/blah/2-files: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./0-many/blah/1-of: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1799: ignoring ./0-many/blah/0-lots: Ignore(IgnoreMatch(Override(Glob(UnmatchedIgnore))))
rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1802: whitelisting ./0-many/GOOD/COOL.log: Whitelist(IgnoreMatch(Override(Glob(Matched(Glob { from: None, original: "*/GOOD/*.log", actual: "*/GOOD/*.log", is_whitelist: false, is_only_dir: false })))))
0-many/GOOD/COOL.log

Same result if I try /*/GOOD/*.log

What is the expected behavior?

Ripgrep should skip recursing into directories that do not match the glob.

Something like this:

$ rg --debug --files -g '/*/GOOD/*.log'
rg: DEBUG|rg::flags::parse|crates/core/flags/parse.rs:97: no extra arguments found from configuration file
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1100: using heuristics to determine whether to read from stdin or search ./ (is_readable_stdin=false, stdin_consumed=false, mode=Files)
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1110: heuristic chose to search ./
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1261: found hostname for hyperlink configuration: timmy
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:1271: hyperlink format: ""
rg: DEBUG|globset|crates/globset/src/lib.rs:453: built glob set; 0 literals, 0 basenames, 0 extensions, 0 prefixes, 0 suffixes, 1 required extensions, 0 regexes
rg: DEBUG|rg::flags::hiargs|crates/core/flags/hiargs.rs:174: using 12 thread(s)
rg: DEBUG|globset|crates/globset/src/lib.rs:453: built glob set; 0 literals, 3 basenames, 0 extensions, 0 prefixes, 0 suffixes, 0 required extensions, 0 regexes
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs: ignoring ./1-dirs/blah/: does not match glob
rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1802: whitelisting ./1-dirs/GOOD/COOL.log: Whitelist(IgnoreMatch(Override(Glob(Matched(Glob { from: None, original: "/*/GOOD/*.log", actual: "*/GOOD/*.log", is_whitelist: false, is_only_dir: false })))))
1-dirs/GOOD/COOL.log
>>>rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs: ignoring ./0-many/blah/: does not match glob
rg: DEBUG|ignore::walk|crates/ignore/src/walk.rs:1802: whitelisting ./0-many/GOOD/COOL.log: Whitelist(IgnoreMatch(Override(Glob(Matched(Glob { from: None, original: "/*/GOOD/*.log", actual: "*/GOOD/*.log", is_whitelist: false, is_only_dir: false })))))
0-many/GOOD/COOL.log
@BurntSushi BurntSushi added the enhancement An enhancement to the functionality of the software. label Apr 25, 2024
@BurntSushi
Copy link
Owner

I thought there was already an open issue for this, but I couldn't find it.

This is a rather difficult optimization to do and is blocked on a rewrite of globset and probably ignore. It's something I've been working on off-and-on for a while now, but it's unlikely to land any time soon. I'm not even 100% certain it's possible unless the --no-ignore flag is also passed. The interaction point between -g/--glob and ignore files will need to be carefully considered for something like this.

Your best bet is to use some other tool to filter out files first. Possibly even using your shell's glob support. Although those could in theory end up being slower than ripgrep even when ripgrep visits more than it needs to. It depends.

@BGR360
Copy link
Author

BGR360 commented Apr 25, 2024

Alright, thanks for clarifying. Using fd to write all of the interesting filepaths to a file and then xargs-ing that into rg got me results way faster!

It would be neat if rg supported piping in filenames from stdin. So I wouldn't have to wait for fd to finish its scan. Should I post another enhancement issue for that?

@BurntSushi
Copy link
Owner

BurntSushi commented Apr 25, 2024

That's #273. But you shouldn't need it. You should be able to pipe the output of fd straight into ripgrep with xargs without writing to an intermediate file.

@BGR360
Copy link
Author

BGR360 commented Apr 25, 2024

I can use xargs yes but ripgrep won't start searching any of those paths until the fd scan completes.

EDIT: oh nvm i can make xargs chunk it up into multiple rg invocations

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement An enhancement to the functionality of the software.
Projects
None yet
Development

No branches or pull requests

2 participants