gc: implement basic GC for Git backend

This adds an initial `jj util gc` command, which simply calls `git gc`
when using the Git backend. That should already be useful in
non-colocated repos because it's not obvious how to GC (repack) such
repos. In my own jj repo, it shrunk `.jj/repo/store/` from 2.4 GiB to
780 MiB, and `jj log --ignore-working-copy` was sped up from 157 ms to
86 ms.

I haven't added any tests because the functionality depends on having
`git` binary on the PATH, which we don't yet depend on anywhere
else. I think we'll still be able to test much of the future parts of
garbage collection without a `git` binary because the interesting
parts are about manipulating the Git repo before calling `git gc` on
it.

CHANGELOG.md

@@ -35,6 +35,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * `jj rebase` now takes the flag `--skip-empty`, which doesn't copy over commits
   that would become empty after a rebase.
 
+* There is a new `jj util gc` command for cleaning up the repository storage.
+  For now, it simply runs `git gc` on the backing Git repo (when using the Git
+  backend).
+
 ### Fixed bugs
 
 

cli/examples/custom-backend/main.rs

@@ -170,4 +170,8 @@ impl Backend for JitBackend {
     ) -> BackendResult<(CommitId, Commit)> {
         self.inner.write_commit(contents, sign_with)
     }
+
+    fn gc(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        self.inner.gc()
+    }
 }

cli/src/commands/util.rs

@@ -15,15 +15,17 @@
 use std::io::Write;
 
 use clap::Subcommand;
+use jj_lib::repo::Repo;
 use tracing::instrument;
 
-use crate::cli_util::{CommandError, CommandHelper};
+use crate::cli_util::{user_error, CommandError, CommandHelper};
 use crate::ui::Ui;
 
 /// Infrequently used commands such as for generating shell completions
 #[derive(Subcommand, Clone, Debug)]
 pub(crate) enum UtilCommand {
     Completion(UtilCompletionArgs),
+    Gc(UtilGcArgs),
     Mangen(UtilMangenArgs),
     ConfigSchema(UtilConfigSchemaArgs),
 }
@@ -56,6 +58,10 @@ pub(crate) struct UtilCompletionArgs {
     zsh: bool,
 }
 
+/// Run backend-dependent garbage collection.
+#[derive(clap::Args, Clone, Debug)]
+pub(crate) struct UtilGcArgs {}
+
 /// Print a ROFF (manpage)
 #[derive(clap::Args, Clone, Debug)]
 pub(crate) struct UtilMangenArgs {}
@@ -84,6 +90,15 @@ pub(crate) fn cmd_util(
             clap_complete::generate(shell, &mut app, "jj", &mut buf);
             ui.stdout_formatter().write_all(&buf)?;
         }
+        UtilCommand::Gc(_gc_args) => {
+            let workspace_command = command.workspace_helper(ui)?;
+            workspace_command
+                .repo()
+                .store()
+                .backend()
+                .gc()
+                .map_err(|err| user_error(err.to_string()))?;
+        }
         UtilCommand::Mangen(_mangen_args) => {
             let mut buf = vec![];
             let man = clap_mangen::Man::new(command.app().clone());

lib/src/backend.rs

@@ -533,4 +533,8 @@ pub trait Backend: Send + Sync + Debug {
         contents: Commit,
         sign_with: Option<&mut SigningFn>,
     ) -> BackendResult<(CommitId, Commit)>;
+
+    /// Perform garbage collection.
+    // TODO: pass in the set of commits to keep here
+    fn gc(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>>;
 }

lib/src/git_backend.rs

@@ -94,6 +94,12 @@ impl From<GitBackendError> for BackendError {
     }
 }
 
+#[derive(Debug, Error)]
+pub enum GitGcError {
+    #[error("Failed to run git gc command: {0}")]
+    GcCommand(std::io::Error),
+}
+
 pub struct GitBackend {
     // While gix::Repository can be created from gix::ThreadSafeRepository, it's
     // cheaper to cache the thread-local instance behind a mutex than creating
@@ -1007,6 +1013,16 @@ impl Backend for GitBackend {
         self.save_extra_metadata_table(mut_table, &table_lock)?;
         Ok((id, contents))
     }
+
+    fn gc(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        let mut git = std::process::Command::new("git");
+        git.env("GIT_DIR", self.git_repo_path());
+        git.args(["gc"]);
+        // TODO: pass output to UI layer instead of printing directly here
+        let mut git_gc = git.spawn().map_err(GitGcError::GcCommand)?;
+        git_gc.wait()?;
+        Ok(())
+    }
 }
 
 /// Write a tree conflict as a special tree with `.jjconflict-base-N` and

lib/src/local_backend.rs

@@ -297,6 +297,10 @@ impl Backend for LocalBackend {
             .map_err(to_other_err)?;
         Ok((id, commit))
     }
+
+    fn gc(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        Ok(())
+    }
 }
 
 pub fn commit_to_proto(commit: &Commit) -> crate::protos::local_store::Commit {

lib/src/store.rs

@@ -67,6 +67,10 @@ impl Store {
         })
     }
 
+    pub fn backend(&self) -> &dyn Backend {
+        self.backend.as_ref()
+    }
+
     pub fn backend_impl(&self) -> &dyn Any {
         self.backend.as_any()
     }

lib/testutils/src/test_backend.rs

@@ -292,4 +292,8 @@ impl Backend for TestBackend {
             .insert(id.clone(), contents.clone());
         Ok((id, contents))
     }
+
+    fn gc(&self) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+        Ok(())
+    }
 }