Merge branch 'add-topo-walk'

Author: Byron

Cargo.lock

@@ -8,7 +8,7 @@ use clap::Parser;
 use gix::{
     bstr::{BString, ByteSlice},
     date::time::format,
-    traverse::commit::Sorting,
+    traverse::commit::simple::Sorting,
 };
 
 fn main() {

examples/log.rs

@@ -175,7 +175,7 @@ where
                         }
                         commit_idx += 1;
                     }
-                    Err(gix::traverse::commit::ancestors::Error::Find { .. }) => {
+                    Err(gix::traverse::commit::simple::Error::Find { .. }) => {
                         is_shallow = true;
                         break;
                     }

gitoxide-core/src/hours/mod.rs

@@ -130,7 +130,7 @@ where
                 .collect::<Result<Vec<_>, _>>()?;
             let handle = repo.objects.into_shared_arc().to_cache_arc();
             let iter = Box::new(
-                traverse::commit::Ancestors::new(tips, traverse::commit::ancestors::State::default(), handle.clone())
+                traverse::commit::Simple::new(tips, handle.clone())
                     .map(|res| res.map_err(|err| Box::new(err) as Box<_>).map(|c| c.id))
                     .inspect(move |_| progress.inc()),
             );
@@ -361,7 +361,7 @@ pub mod input_iteration {
     #[derive(Debug, thiserror::Error)]
     pub enum Error {
         #[error("input objects couldn't be iterated completely")]
-        Iteration(#[from] traverse::commit::ancestors::Error),
+        Iteration(#[from] traverse::commit::simple::Error),
         #[error("An error occurred while reading hashes from standard input")]
         InputLinesIo(#[from] std::io::Error),
         #[error("Could not decode hex hash provided on standard input")]

gitoxide-core/src/pack/create.rs

@@ -429,7 +429,7 @@ pub fn update(
                         break;
                     }
                 }
-                Err(gix::traverse::commit::ancestors::Error::Find { .. }) => {
+                Err(gix::traverse::commit::simple::Error::Find { .. }) => {
                     writeln!(err, "shallow repository - commit history is truncated").ok();
                     break;
                 }

gitoxide-core/src/query/engine/update.rs

@@ -2,7 +2,7 @@ pub(crate) mod function {
     use std::{borrow::Cow, ffi::OsString};
 
     use anyhow::{bail, Context};
-    use gix::{prelude::ObjectIdExt, traverse::commit::Sorting};
+    use gix::{prelude::ObjectIdExt, traverse::commit::simple::Sorting};
 
     use crate::OutputFormat;
 

gitoxide-core/src/repository/commitgraph/list.rs

@@ -17,7 +17,7 @@ pub const PROGRESS_RANGE: std::ops::RangeInclusive<u8> = 0..=2;
 
 pub(crate) mod function {
     use anyhow::{bail, Context};
-    use gix::{hashtable::HashMap, traverse::commit::Sorting, Progress};
+    use gix::{hashtable::HashMap, traverse::commit::simple::Sorting, Progress};
     use layout::{
         backends::svg::SVGWriter,
         core::{base::Orientation, geometry::Point, style::StyleAttr},

gitoxide-core/src/repository/revision/list.rs

@@ -42,7 +42,7 @@ impl<'a> TryFrom<&'a BStr> for File<'a> {
 
 impl From<File<'_>> for BString {
     fn from(c: File<'_>) -> Self {
-        c.into()
+        c.to_bstring()
     }
 }
 

gix-config/src/file/impls.rs

@@ -39,7 +39,7 @@ impl Display for Comment<'_> {
 
 impl From<Comment<'_>> for BString {
     fn from(c: Comment<'_>) -> Self {
-        c.into()
+        c.to_bstring()
     }
 }
 

gix-config/src/parse/comment.rs

@@ -72,7 +72,7 @@ impl Display for Event<'_> {
 
 impl From<Event<'_>> for BString {
     fn from(event: Event<'_>) -> Self {
-        event.into()
+        event.to_bstring()
     }
 }
 

gix-config/src/parse/event.rs

@@ -145,7 +145,7 @@ impl Display for Header<'_> {
 
 impl From<Header<'_>> for BString {
     fn from(header: Header<'_>) -> Self {
-        header.into()
+        header.to_bstring()
     }
 }
 

gix-config/src/parse/section/header.rs

@@ -129,11 +129,10 @@ mod changes {
         }
 
         fn all_commits(db: &gix_odb::Handle) -> HashMap<String, ObjectId> {
-            use gix_traverse::commit;
             let mut buf = Vec::new();
 
             let head = head_of(db);
-            commit::Ancestors::new(Some(head), commit::ancestors::State::default(), &db)
+            gix_traverse::commit::Simple::new(Some(head), &db)
                 .collect::<Result<Vec<_>, _>>()
                 .expect("valid iteration")
                 .into_iter()

gix-diff/tests/tree/mod.rs

@@ -57,12 +57,14 @@ impl<T> Tree<T> {
             })?,
         );
 
-        let anticipated_num_objects = if let Some(num_objects) = data_sorted_by_offsets.size_hint().1 {
-            progress.init(Some(num_objects), progress::count("objects"));
-            num_objects
-        } else {
-            0
-        };
+        let anticipated_num_objects = data_sorted_by_offsets
+            .size_hint()
+            .1
+            .map(|num_objects| {
+                progress.init(Some(num_objects), progress::count("objects"));
+                num_objects
+            })
+            .unwrap_or_default();
         let mut tree = Tree::with_capacity(anticipated_num_objects)?;
 
         {

gix-pack/src/cache/delta/from_offsets.rs

@@ -9,7 +9,6 @@ use gix_pack::data::{
     output,
     output::{count, entry},
 };
-use gix_traverse::commit;
 
 use crate::pack::{
     data::output::{db, DbKind},
@@ -241,7 +240,7 @@ fn traversals() -> crate::Result {
         .copied()
         {
             let head = hex_to_id("dfcb5e39ac6eb30179808bbab721e8a28ce1b52e");
-            let mut commits = commit::Ancestors::new(Some(head), commit::ancestors::State::default(), db.clone())
+            let mut commits = gix_traverse::commit::Simple::new(Some(head), db.clone())
                 .map(Result::unwrap)
                 .map(|c| c.id)
                 .collect::<Vec<_>>();

gix-pack/tests/pack/data/output/count_and_entries.rs

@@ -22,3 +22,4 @@ gix-revwalk = { version = "^0.13.0", path = "../gix-revwalk" }
 gix-commitgraph = { version = "^0.24.2", path = "../gix-commitgraph" }
 smallvec = "1.10.0"
 thiserror = "1.0.32"
+bitflags = "2"

gix-traverse/Cargo.toml

@@ -0,0 +1,88 @@
+//! Provide multiple traversal implementations with different performance envelopes.
+//!
+//! Use [`Simple`] for fast walks that maintain minimal state, or [`Topo`] for a more elaborate traversal.
+use gix_hash::ObjectId;
+use gix_object::FindExt;
+use gix_revwalk::graph::IdMap;
+use gix_revwalk::PriorityQueue;
+use smallvec::SmallVec;
+
+/// A fast iterator over the ancestors of one or more starting commits.
+pub struct Simple<Find, Predicate> {
+    objects: Find,
+    cache: Option<gix_commitgraph::Graph>,
+    predicate: Predicate,
+    state: simple::State,
+    parents: Parents,
+    sorting: simple::Sorting,
+}
+
+/// Simple ancestors traversal, without the need to keep track of graph-state.
+pub mod simple;
+
+/// A commit walker that walks in topographical order, like `git rev-list
+/// --topo-order` or `--date-order` depending on the chosen [`topo::Sorting`].
+///
+/// Instantiate with [`topo::Builder`].
+pub struct Topo<Find, Predicate> {
+    commit_graph: Option<gix_commitgraph::Graph>,
+    find: Find,
+    predicate: Predicate,
+    indegrees: IdMap<i32>,
+    states: IdMap<topo::WalkFlags>,
+    explore_queue: PriorityQueue<topo::iter::GenAndCommitTime, ObjectId>,
+    indegree_queue: PriorityQueue<topo::iter::GenAndCommitTime, ObjectId>,
+    topo_queue: topo::iter::Queue,
+    parents: Parents,
+    min_gen: u32,
+    buf: Vec<u8>,
+}
+
+pub mod topo;
+
+/// Specify how to handle commit parents during traversal.
+#[derive(Default, Copy, Clone)]
+pub enum Parents {
+    /// Traverse all parents, useful for traversing the entire ancestry.
+    #[default]
+    All,
+    /// Only traverse along the first parent, which commonly ignores all branches.
+    First,
+}
+
+/// The collection of parent ids we saw as part of the iteration.
+///
+/// Note that this list is truncated if [`Parents::First`] was used.
+pub type ParentIds = SmallVec<[gix_hash::ObjectId; 1]>;
+
+/// Information about a commit that we obtained naturally as part of the iteration.
+#[derive(Debug, Clone, PartialEq, Eq, Ord, PartialOrd, Hash)]
+pub struct Info {
+    /// The id of the commit.
+    pub id: gix_hash::ObjectId,
+    /// All parent ids we have encountered. Note that these will be at most one if [`Parents::First`] is enabled.
+    pub parent_ids: ParentIds,
+    /// The time at which the commit was created. It will only be `Some(_)` if the chosen traversal was
+    /// taking dates into consideration.
+    pub commit_time: Option<gix_date::SecondsSinceUnixEpoch>,
+}
+
+enum Either<'buf, 'cache> {
+    CommitRefIter(gix_object::CommitRefIter<'buf>),
+    CachedCommit(gix_commitgraph::file::Commit<'cache>),
+}
+
+fn find<'cache, 'buf, Find>(
+    cache: Option<&'cache gix_commitgraph::Graph>,
+    objects: Find,
+    id: &gix_hash::oid,
+    buf: &'buf mut Vec<u8>,
+) -> Result<Either<'buf, 'cache>, gix_object::find::existing_iter::Error>
+where
+    Find: gix_object::Find,
+{
+    match cache.and_then(|cache| cache.commit_by_id(id).map(Either::CachedCommit)) {
+        Some(c) => Ok(c),
+        None => objects.find_commit_iter(id, buf).map(Either::CommitRefIter),
+    }
+}