Read a large change once for all of a checkout's batches 519c7c061bcb

Building graphs a batch at a time, a checkout of the Linux tree read the
81 MiB first change again for every batch of 2 048 files, for the lines
of its inline inserts. The repository now keeps the last change it read
for that, shared, so every batch after the first finds it. A local
clone of the Linux tip takes 56.8 s (from 61 s); its checkout 26.4 s.

refs #35

Author: Andy McCall @andymccall <[email protected]>
When: 1 month ago · 2026-09-29 23:55 UTC
Signature: ✔ signed by Andy McCall
Hash: 519c7c061bcb7bc33441cc09f32403d048ed21bdf12f413d4c109b38e336853a
Depends on: 1d360c91a289 598009a75f6c 75167672ebcd 982864068a59 a4fad5b085c7 c863c7a7a2d1 cc7a00565161 e8d38e47e9d7 ff0af1ee299d
Depended on by: c895bc52e4e0 f53cbe7a204e 98a1620a3840 acadb5ed28a1 57d8dc05c9d4 2c75672fb2fe
In bookmarks: agents-in-readme bare-hashes carried-not-removed carry-identical doctor-trust identical-sides keyless-tags main merge-remote-bookmark only-help-103 paths-for-resolutions-and-amend push-notes show-stat subrepo-pin-state switch-says
In tags: v0.1.0 v0.10.0 v0.11.0 v0.2.0 v0.3.0 v0.3.1 v0.4.0 v0.4.1 v0.4.2 v0.4.3 v0.4.4 v0.4.5 v0.4.6 v0.5.0 v0.5.1 v0.6.0 v0.7.0 v0.8.0 v0.9.0
Edits: 14 atoms, 0 tree operations, 3 files

Browse the tree at this change

hide whitespace changes

modified crates/oxy-core/src/repo/mod.rs +19 −4

modified crates/oxy-core/src/state.rs +3 −3

modified docs/benchmarks.md +1 −0

		…	178 unchanged lines
179	179		pub config: Config,
180	180		/// What the working copy's file system can represent (spec §7.7).
181	181		platform: crate::portable::Platform,
	182	+	/// The encoding of the change last read for its inline inserts.
	183	+	last_change: std::cell::RefCell<Option<(ChangeHash, std::sync::Arc<Vec<u8>>)>>,
182	184		}
183	185
184	186		/// Marker file of a bare repository, inside `.oxy/`.
		…	59 unchanged lines
244	246		write_atomic(&oxy.join("workspace"), workspace.to_hex().as_bytes())?;
245	247		let store = Store::open(oxy.join("objects"))?;
246	248		let repo = Repo {
	249	+	last_change: Default::default(),
247	250		git_export_skipped: Default::default(),
248	251		root,
249	252		oxy,
		…	43 unchanged lines
293	296		config.save(&oxy.join("config"))?;
294	297		let store = Store::open(oxy.join("objects"))?;
295	298		let repo = Repo {
	299	+	last_change: Default::default(),
296	300		git_export_skipped: Default::default(),
297	301		root,
298	302		oxy,
		…	65 unchanged lines
364	368		let config = Config::load(Some(&default_user_config()), &shared.join("config"))?;
365	369		let store = Store::open(shared.join("objects"))?;
366	370		let mut repo = Repo {
	371	+	last_change: Default::default(),
367	372		git_export_skipped: Default::default(),
368	373		root: d.to_path_buf(),
369	374		oxy: shared,
		…	20 unchanged lines
390	395		};
391	396		let store = Store::open(oxy.join("objects"))?;
392	397		let mut repo = Repo {
	398	+	last_change: Default::default(),
393	399		git_export_skipped: Default::default(),
394	400		root: d.to_path_buf(),
395	401		oxy,
		…	452 unchanged lines
848	854		)?)
849	855		}
850	856
851		-	/// A change's encoding, for building graphs of its inline inserts.
	857	+	/// A change's encoding, for building graphs of its inline inserts. The
	858	+	/// last one read is kept: a checkout building graphs a batch at a time
	859	+	/// reads the same large change for every batch.
852	860		fn change_bytes(
853	861		&self,
854	862		h: &ChangeHash,
855		-	) -> std::result::Result<Vec<u8>, crate::objects::ObjectError> {
856		-	self.store.get_typed(h, ObjectType::Change).map_err(\|e\| {
	863	+	) -> std::result::Result<std::sync::Arc<Vec<u8>>, crate::objects::ObjectError> {
	864	+	if let Some((k, b)) = &*self.last_change.borrow() {
	865	+	if k == h {
	866	+	return Ok(b.clone());
	867	+	}
	868	+	}
	869	+	let b = std::sync::Arc::new(self.store.get_typed(h, ObjectType::Change).map_err(\|e\| {
857	870		crate::objects::ObjectError::Schema(format!("change {}: {e}", h.short(12)))
858		-	})
	871	+	})?);
	872	+	self.last_change.borrow_mut() = Some((h, b.clone()));
	873	+	Ok(b)
859	874		}
860	875
861	876		/// Load the dependency closure of `heads` change by change.
		…	853 unchanged lines

		…	281 unchanged lines
282	282		&mut self,
283	283		nodes: &[NodeId],
284	284		load: impl FnOnce(&[ObjectRef]) -> Result<Vec<Vec<u8>>, ObjectError>,
285		-	read_change: impl Fn(&ChangeHash) -> Result<Vec<u8>, ObjectError>,
	285	+	read_change: impl Fn(&ChangeHash) -> Result<Arc<Vec<u8>>, ObjectError>,
286	286		) -> Result<(), ObjectError> {
287	287		let Some(deferred) = self.deferred.as_mut() else {
288	288		return Ok(());
		…	19 unchanged lines
308	308		}));
309	309		// And the encodings of changes whose inline inserts are needed,
310	310		// each read once.
311		-	let mut changes: HashMap<ChangeHash, Vec<u8>> = HashMap::new();
	311	+	let mut changes: HashMap<ChangeHash, Arc<Vec<u8>>> = HashMap::new();
312	312		for t in taken.iter().flat_map(\|(_, d)\| d.texts.iter()) {
313	313		if let TextRef::Inline { change, .. } = t {
314	314		if !changes.contains_key(change) {
		…	103 unchanged lines
418	418		pub fn load_all_graphs(
419	419		&mut self,
420	420		load: impl FnOnce(&[ObjectRef]) -> Result<Vec<Vec<u8>>, ObjectError>,
421		-	read_change: impl Fn(&ChangeHash) -> Result<Vec<u8>, ObjectError>,
	421	+	read_change: impl Fn(&ChangeHash) -> Result<Arc<Vec<u8>>, ObjectError>,
422	422		) -> Result<(), ObjectError> {
423	423		let nodes: Vec<NodeId> = match &self.deferred {
424	424		Some(d) => d.keys().copied().collect(),
		…	1499 unchanged lines

		…	409 unchanged lines
410	410		\|---\|---:\|---:\|
411	411		\| Before \| 96 s (81.6 s in a quieter run) \| 5.2 to 5.4 GB \|
412	412		\| After \| 61.0 s \| 4.1 GB \|
	413	+	\| The change read once for all batches \| 56.8 s \| 4.7 GB \|
413	414		\| Git \| 9.6 s \| 469 MB \|
414	415
415	416		The checkout (31 s) now spends most of its time building graphs: each
		…	3 unchanged lines

`/`	Search
`g` `c`	Code
`g` `h`	History
`g` `b`	Bookmarks
`g` `t`	Trust
`y`	Permalink to this state (files and trees)
`?`	This sheet

Read a large change once for all of a checkout's batches 519c7c061bcb

Keyboard