Skip to content

Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows

refactor(perf): canonicalize wt-perf prune-fixture layering + completion config (#3548)

Maximilian Roos166e1190ba55
diff --git a/benches/CLAUDE.md b/benches/CLAUDE.md
index 5e4cc2dd4..7538c77ce 100644
--- a/benches/CLAUDE.md
+++ b/benches/CLAUDE.md
@@ -32,6 +32,32 @@ cargo bench --bench picker_preview # all variants
cargo bench --bench picker_preview warm # warm only
```
+## Fixtures and Benches
+
+**Bench groups are named for what they measure; fixtures for the repo shape they
+build.** The two never match by convention because the relationship is
+many-to-one — one fixture feeds several bench groups that measure different
+things (`RepoConfig::typical` alone backs five). Don't read a bench group and its
+fixture as duplicates of each other: `full` is a benchmark, `mixed` is the repo
+it runs on.
+
+| Fixture (generator) | `wt-perf setup` handle | Bench group(s) |
+|---|---|---|
+| `RepoConfig::typical(N)` | `typical-N` | `skeleton`, `worktree_scaling` (list.rs); `first_output` (time_to_first_output.rs); `picker_preview`; `remove_e2e` |
+| `RepoConfig::branches(N, M)` | `branches-N[-M]` | `completion_switch` (completion.rs) |
+| `RepoConfig::many_divergent_branches()` | `divergent` | `divergent_branches` (list.rs) |
+| `create_mixed_repo(W, B)` | `mixed-W-B` | `full` (list.rs) |
+| `create_prune_repo_at(M, U)` | `prune-M-U` | `prune_e2e` (prune.rs) |
+| `ensure_prune_real_repo(M, U)` | `prune-real[-M-U]` | `prune_real_repo` (prune.rs) |
+| rust-lang/rust clone (`clone_rust_repo`) | — | `real_repo`, `real_repo_many_branches` (list.rs) |
+| `lean_worktrees` (local to alias.rs) | — | `dispatch` (alias.rs) |
+| `RepoConfig::picker_test()` | `picker-test` | — (interactive picker debugging only) |
+
+Renaming a bench group is not free: `.github/scripts/criterion-to-jsonl.py`
+keys each time-series row by the criterion output path (`<group>/<id>`), and the
+daily `benchmarks` workflow appends those to a gist. A rename orphans that
+series.
+
## Rust Repo Caching
Real repo benchmarks clone rust-lang/rust on first run (~2-5 minutes). The clone is cached in `target/wt-perf/bench-repos/` and reused. Corrupted caches are auto-recovered.
diff --git a/benches/completion.rs b/benches/completion.rs
index 6c9437b72..f10660c1e 100644
--- a/benches/completion.rs
+++ b/benches/completion.rs
@@ -23,15 +23,7 @@ fn bench_completion_switch(c: &mut Criterion) {
// Without worktrees: all branches are candidates
group.bench_function("branches_only", |b| {
- let config = RepoConfig {
- commits_on_main: 1,
- files: 1,
- branches: 50,
- commits_per_branch: 0,
- worktrees: 0,
- worktree_commits_ahead: 0,
- worktree_uncommitted_files: 0,
- };
+ let config = RepoConfig::branches(50, 0);
let temp = create_repo(&config);
let repo = temp.path().join("repo");
b.iter(|| run_completion(binary, &repo, &["wt", "switch", ""]));
@@ -40,13 +32,8 @@ fn bench_completion_switch(c: &mut Criterion) {
// With worktrees: filters out branches that already have worktrees
group.bench_function("with_worktrees", |b| {
let config = RepoConfig {
- commits_on_main: 1,
- files: 1,
- branches: 50,
- commits_per_branch: 0,
worktrees: 10,
- worktree_commits_ahead: 0,
- worktree_uncommitted_files: 0,
+ ..RepoConfig::branches(50, 0)
};
let temp = create_repo(&config);
let repo = temp.path().join("repo");
diff --git a/tests/helpers/wt-perf/src/lib.rs b/tests/helpers/wt-perf/src/lib.rs
index 60d874826..3bc2a4408 100644
--- a/tests/helpers/wt-perf/src/lib.rs
+++ b/tests/helpers/wt-perf/src/lib.rs
@@ -1119,8 +1119,7 @@ pub fn create_prune_repo_at(merged: usize, unmerged: usize, base_path: &Path) {
worktree_uncommitted_files: 0,
};
create_repo_at(&config, base_path);
- add_diverged_backdrop(base_path, unmerged, unmerged);
- add_squash_merged(base_path, merged, 0);
+ add_prune_populations(base_path, merged, unmerged);
// The squash commits advanced main past the fake remote ref written by
// `create_repo_at`; refresh so origin/main tracks the final tip.
setup_fake_remote(base_path);
@@ -1209,6 +1208,20 @@ pub fn add_squash_merged(repo_path: &Path, count: usize, round: usize) {
}
}
+/// Add the two populations that turn a repo into a `wt step prune` workload:
+/// `unmerged` two-sided-diverged worktrees and branches (`add_diverged_backdrop`
+/// — the backdrop prune scans every run but never removes) and `merged`
+/// squash-merged candidate pairs (`add_squash_merged` — what prune removes).
+///
+/// The base repo is the only thing the synthetic ([`create_prune_repo_at`]) and
+/// rust-scale (`create_prune_real_repo_at`) prune fixtures differ in; both layer
+/// these identical populations on top, so keeping the layering in one place
+/// stops the two fixtures from drifting.
+fn add_prune_populations(base_path: &Path, merged: usize, unmerged: usize) {
+ add_diverged_backdrop(base_path, unmerged, unmerged);
+ add_squash_merged(base_path, merged, 0);
+}
+
/// Default populations for the rust-scale prune fixture (`prune-real`):
/// 12 squash-merged candidates of each kind + 24 unmerged worktrees and
/// branches → 36 linked worktrees, and a live prune that removes 24
@@ -1236,8 +1249,7 @@ pub const PRUNE_REAL_UNMERGED: usize = 24;
/// `target/wt-perf/bench-repos` and repairs consumed candidates on later runs.
fn create_prune_real_repo_at(merged: usize, unmerged: usize, base_path: &Path) {
clone_rust_repo_at(base_path);
- add_diverged_backdrop(base_path, unmerged, unmerged);
- add_squash_merged(base_path, merged, 0);
+ add_prune_populations(base_path, merged, unmerged);
}
/// How a cached prune fixture compares to its expected populations.