feat: `revision::walk::Platform::selected(filter)` to selectively prune parts of the commit graph. (#801)

Sebastian Thiel committed 3 years ago

Commit ec93f751 parent a70085e2 files · +67 −21Browse files

Unified · Split

2 files

gix/src/revision/walk.rs +36 −21
gix/tests/id/mod.rs +31 −0

 /// Produce the iterator
 impl<'repo> Platform<'repo> {
-    /// Return an iterator to traverse all commits reachable as configured by the [Platform].
+    /// For each commit, let `filter` return `true` if it and its parents should be included in the traversal, or `false`
-    ///
+    /// if the traversal should exclude it and its ancestry entirely.
-    /// # Performance
     ///
-    /// It's highly recommended to set an [`object cache`][Repository::object_cache_size()] on the parent repo
+    /// If `filter` is None, no pruning of the graph will be performed which is the default.
-    /// to greatly speed up performance if the returned id is supposed to be looked up right after.
+    pub fn selected(
-    pub fn all(self) -> Result<revision::Walk<'repo>, Error> {
+        self,
+        mut filter: impl FnMut(&gix_hash::oid) -> bool + 'repo,
+    ) -> Result<revision::Walk<'repo>, Error> {
         let Platform {
             repo,
             tips,
                         let shallow_commits = repo.shallow_commits()?;
                         let mut grafted_parents_to_skip = Vec::new();
                         let mut buf = Vec::new();
-                        move |id| match shallow_commits.as_ref() {
+                        move |id| {
-                            Some(commits) => {
+                            if !filter(id) {
-                                let id = id.to_owned();
+                                return false;
-                                if let Ok(idx) = grafted_parents_to_skip.binary_search(&id) {
+                            }
-                                    grafted_parents_to_skip.remove(idx);
+                            match shallow_commits.as_ref() {
-                                    return false;
+                                Some(commits) => {
-                                };
+                                    let id = id.to_owned();
-                                if commits.binary_search(&id).is_ok() {
+                                    if let Ok(idx) = grafted_parents_to_skip.binary_search(&id) {
-                                    if let Ok(commit) = repo.objects.find_commit_iter(id, &mut buf) {
+                                        grafted_parents_to_skip.remove(idx);
-                                        grafted_parents_to_skip.extend(commit.parent_ids());
+                                        return false;
-                                        grafted_parents_to_skip.sort();
+                                    };
-                                    }
+                                    if commits.binary_search(&id).is_ok() {
-                                };
+                                        if let Ok(commit) = repo.objects.find_commit_iter(id, &mut buf) {
-                                true
+                                            grafted_parents_to_skip.extend(commit.parent_ids());
+                                            grafted_parents_to_skip.sort();
+                                        }
+                                    };
+                                    true
+                                }
+                                None => true,
                             }
-                            None => true,
                         }
                     },
                 )
             ),
         })
     }
+    /// Return an iterator to traverse all commits reachable as configured by the [Platform].
+    ///
+    /// # Performance
+    ///
+    /// It's highly recommended to set an [`object cache`][Repository::object_cache_size()] on the parent repo
+    /// to greatly speed up performance if the returned id is supposed to be looked up right after.
+    pub fn all(self) -> Result<revision::Walk<'repo>, Error> {
+        self.selected(|_| true)
+    }
 }
 pub(crate) mod iter {

gix/tests/id/mod.rs+31−0

@@ -67,6 +67,7 @@ fn display_and_debug() -> crate::Result { · 66 unmodified lines
67	}	67	}
68		68
69	mod ancestors {	69	mod ancestors {
		70	use crate::id::hex_to_id;
70	use gix_traverse::commit;	71	use gix_traverse::commit;
71		72
72	#[test]	73	#[test]
@@ -98,4 +99,34 @@ mod ancestors { · 25 unmodified lines
98	);	99	);
99	Ok(())	100	Ok(())
100	}	101	}
		102
		103	#[test]
		104	fn filtered() -> crate::Result {
		105	let repo = crate::repo("make_repo_with_fork_and_dates.sh")?.to_thread_local();
		106	let head = repo.head()?.into_fully_peeled_id().expect("born")?;
		107
		108	for sorting in [
		109	commit::Sorting::Topological,
		110	commit::Sorting::ByCommitTimeNewestFirst,
		111	commit::Sorting::ByCommitTimeNewestFirstCutoffOlderThan {
		112	time_in_seconds_since_epoch: 0,
		113	},
		114	] {
		115	let commits_graph_order = head
		116	.ancestors()
		117	.sorting(sorting)
		118	.selected(\|id\| {
		119	let _assert_lifetime_works = &repo; // assure we can use repo here.
		120	id != hex_to_id("9902e3c3e8f0c569b4ab295ddf473e6de763e1e7")
		121	&& id != hex_to_id("bcb05040a6925f2ff5e10d3ae1f9264f2e8c43ac")
		122	})?
		123	.collect::<Result<Vec<_>, _>>()?;
		124	assert_eq!(
		125	commits_graph_order,
		126	&[hex_to_id("288e509293165cb5630d08f4185bdf2445bf6170")],
		127	"we ignore all but the first"
		128	);
		129	}
		130	Ok(())
		131	}
101	}	132	}