girt

Read a repository

Print the first ten commits in HEAD’s history. The getting started page builds this program step by step.

src/main.rs

use girt::{HistoryLimits, PackLimits, ReadLimits, Repository};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let repository = Repository::open(".")?;

    let head = repository.head()?;
    let head_id = head.id.ok_or("HEAD has no commits yet")?;

    let objects = repository.objects()?;
    let history = objects.walk(&[head_id])?;

    for id in history.into_iter().take(10) {
        let commit = objects.read_commit(id)?;
        let message = String::from_utf8_lossy(commit.message());
        let summary = message.lines().next().unwrap_or_default();
        println!("{id} {summary}");
    }

    Ok(())
}

jj on girt compared with stock jj

girt’s first consumer is jj. This table runs the same jj commands on girt and on stock jj, which uses gix and the git executable, with the closest git command for reference.

	Median time			girt vs stock jj
Command	girt	stock jj	Git	Ratio	Bar
jj status	560 ms	420 ms	9.63 s	1.3× slower
jj log	203 ms	210 ms	5.64 s	about the same
jj diff	710 ms	472 ms	15.5 s	1.5× slower
jj commit	2.12 s	1.13 s	11.7 s	1.9× slower
jj describe	1.60 s	951 ms	9.89 s	1.7× slower
jj rebase	1.70 s	1.07 s	423 ms	1.6× slower
jj git fetch	850 ms	985 ms	744 ms	1.2× faster
jj git push	1.81 s	1.77 s	1.44 s	about the same
jj git clone	—	4 min 53 s	4 min 53 s	times out

Everyday commands on the Linux v6.1 kernel repository. Across all three repositories, jj on girt is less than 20% slower than stock jj (or faster) in 109 of the 152 cells where both finished; 1 more did not finish with girt. Apple M2 Max, girt 0.11.1–0.14.0, measured 2026-10-17. All results

Biggest gaps, by extra wait over stock jj: jj git clone on Linux v6.1 times out (issue 581); jj file annotate on Linux v6.1 1.1× slower; jj log on Linux v6.1 about the same. Performance page

• OverviewWhat girt is, who it's for and where it stands.
• Supported and refusedWhat each area supports and what it rejects.
• jj on girt benchmarkEvery measured command, grouped by how often you run it.