Merge pull request GitoxideLabs#1453 from cruessler/gix-blame

Explore gix APIs, experiment with gix-blame API

README.md

@@ -139,6 +139,7 @@ is usable to some extent.
   * [gix-shallow](https://github.com/GitoxideLabs/gitoxide/blob/main/crate-status.md#gix-shallow)
   * `gitoxide-core`
 * **very early**  _(possibly without any documentation and many rough edges)_
+  * [gix-blame](https://github.com/GitoxideLabs/gitoxide/blob/main/crate-status.md#gix-blame)
 * **idea** _(just a name placeholder)_
   * [gix-note](https://github.com/GitoxideLabs/gitoxide/blob/main/crate-status.md#gix-note)
   * [gix-fetchhead](https://github.com/GitoxideLabs/gitoxide/blob/main/crate-status.md#gix-fetchhead)

crate-status.md

@@ -293,7 +293,7 @@ The top-level crate that acts as hub to all functionality provided by the `gix-*
     * [x] safe with cycles and recursive configurations
     * [x] multi-line with comments and quotes
 * **promisor**
-    * It's vague, but these seems to be like index files allowing to fetch objects from a server on demand.
+    * It's vague, but these seem to be like index files allowing to fetch objects from a server on demand.
 * [x] API documentation
     * [ ] Some examples
 
@@ -361,6 +361,25 @@ Check out the [performance discussion][gix-diff-performance] as well.
 * [x] API documentation
     * [ ] Examples
 
+### gix-blame
+
+* [x] commit-annotations for a single file
+    - [ ] progress
+    - [ ] interruptibility
+    - [ ] streaming
+- [ ] support for worktree changes (creates virtual commit on top of `HEAD`) 
+- [ ] shallow-history support
+- [ ] rename tracking (track different paths through history)
+- [ ] commits to ignore
+- [ ] pass all blame-cornercases (from Git)
+* **Performance-Improvements**
+    * Without the following the performance isn't competitive with Git.
+    1. Implement custom graph walk which won't run down parents that don't have the path in question.
+    2. Implement access of trees from commit-graph and fill that information into the traversal info by default.
+    3. commit-graph with bloom filter, used to quickly check if a commit has a path.
+* [x] API documentation
+    * [ ] Examples
+
 ### gix-traverse
 
 Check out the [performance discussion][gix-traverse-performance] as well.

gitoxide-core/Cargo.toml

@@ -49,7 +49,7 @@ serde = ["gix/serde", "dep:serde_json", "dep:serde", "bytesize/serde"]
 
 [dependencies]
 # deselect everything else (like "performance") as this should be controllable by the parent application.
-gix = { version = "^0.69.1", path = "../gix", default-features = false, features = ["merge", "blob-diff", "revision", "mailmap", "excludes", "attributes", "worktree-mutation", "credentials", "interrupt", "status", "dirwalk"] }
+gix = { version = "^0.69.1", path = "../gix", default-features = false, features = ["merge", "blob-diff", "blame", "revision", "mailmap", "excludes", "attributes", "worktree-mutation", "credentials", "interrupt", "status", "dirwalk"] }
 gix-pack-for-configuration-only = { package = "gix-pack", version = "^0.56.0", path = "../gix-pack", default-features = false, features = ["pack-cache-lru-dynamic", "pack-cache-lru-static", "generate", "streaming-input"] }
 gix-transport-configuration-only = { package = "gix-transport", version = "^0.44.0", path = "../gix-transport", default-features = false }
 gix-archive-for-configuration-only = { package = "gix-archive", version = "^0.18.0", path = "../gix-archive", optional = true, features = ["tar", "tar_gz"] }

gitoxide-core/src/repository/blame.rs

@@ -0,0 +1,71 @@
+use gix::bstr::ByteSlice;
+use gix::config::tree;
+use std::ffi::OsStr;
+
+pub fn blame_file(
+    mut repo: gix::Repository,
+    file: &OsStr,
+    out: impl std::io::Write,
+    err: Option<&mut dyn std::io::Write>,
+) -> anyhow::Result<()> {
+    {
+        let mut config = repo.config_snapshot_mut();
+        if config.string(&tree::Core::DELTA_BASE_CACHE_LIMIT).is_none() {
+            config.set_value(&tree::Core::DELTA_BASE_CACHE_LIMIT, "100m")?;
+        }
+    }
+    let index = repo.index_or_empty()?;
+    repo.object_cache_size_if_unset(repo.compute_object_cache_size_for_tree_diffs(&index));
+
+    let file = gix::path::os_str_into_bstr(file)?;
+    let specs = repo.pathspec(
+        false,
+        [file],
+        true,
+        &index,
+        gix::worktree::stack::state::attributes::Source::WorktreeThenIdMapping.adjust_for_bare(repo.is_bare()),
+    )?;
+    // TODO: there should be a way to normalize paths without going through patterns, at least in this case maybe?
+    //       `Search` actually sorts patterns by excluding or not, all that can lead to strange results.
+    let file = specs
+        .search()
+        .patterns()
+        .map(|p| p.path().to_owned())
+        .next()
+        .expect("exactly one pattern");
+
+    let suspect = repo.head()?.peel_to_commit_in_place()?;
+    let traverse =
+        gix::traverse::commit::topo::Builder::from_iters(&repo.objects, [suspect.id], None::<Vec<gix::ObjectId>>)
+            .with_commit_graph(repo.commit_graph_if_enabled()?)
+            .build()?;
+    let mut resource_cache = repo.diff_resource_cache_for_tree_diff()?;
+    let outcome = gix::blame::file(&repo.objects, traverse, &mut resource_cache, file.as_bstr())?;
+    let statistics = outcome.statistics;
+    write_blame_entries(out, outcome)?;
+
+    if let Some(err) = err {
+        writeln!(err, "{statistics:#?}")?;
+    }
+    Ok(())
+}
+
+fn write_blame_entries(mut out: impl std::io::Write, outcome: gix::blame::Outcome) -> Result<(), std::io::Error> {
+    for (entry, lines_in_hunk) in outcome.entries_with_lines() {
+        for ((actual_lno, source_lno), line) in entry
+            .range_in_blamed_file()
+            .zip(entry.range_in_source_file())
+            .zip(lines_in_hunk)
+        {
+            write!(
+                out,
+                "{short_id} {line_no} {src_line_no} {line}",
+                line_no = actual_lno + 1,
+                src_line_no = source_lno + 1,
+                short_id = entry.commit_id.to_hex_with_len(8),
+            )?;
+        }
+    }
+
+    Ok(())
+}

gitoxide-core/src/repository/mod.rs

@@ -21,6 +21,7 @@ pub enum PathsOrPatterns {
 pub mod archive;
 pub mod cat;
 pub use cat::function::cat;
+pub mod blame;
 pub mod commit;
 pub mod config;
 mod credential;

gix-blame/Cargo.toml

@@ -5,7 +5,7 @@ name = "gix-blame"
 version = "0.0.0"
 repository = "https://github.com/GitoxideLabs/gitoxide"
 license = "MIT OR Apache-2.0"
-description = "A crate of the gitoxide project dedicated implementing a 'blame' algorithm"
+description = "A crate of the gitoxide project dedicated to implementing a 'blame' algorithm"
 authors = ["Christoph Rüßler <[email protected]>", "Sebastian Thiel <[email protected]>"]
 edition = "2021"
 rust-version = "1.65"
@@ -14,6 +14,19 @@ rust-version = "1.65"
 doctest = false
 
 [dependencies]
+gix-trace = { version = "^0.1.11", path = "../gix-trace" }
+gix-diff = { version = "^0.49.0", path = "../gix-diff", default-features = false, features = ["blob"] }
+gix-object = { version = "^0.46.0", path = "../gix-object" }
+gix-hash = { version = "^0.15.0", path = "../gix-hash" }
+gix-worktree = { version = "^0.38.0", path = "../gix-worktree", default-features = false, features = ["attributes"] }
+gix-traverse = { version = "^0.43.0", path = "../gix-traverse" }
+
+thiserror = "2.0.0"
 
 [dev-dependencies]
+gix-ref = { version = "^0.49.0", path = "../gix-ref" }
+gix-filter = { version = "^0.16.0", path = "../gix-filter" }
+gix-fs = { version = "^0.12.0", path = "../gix-fs" }
+gix-index = { version = "^0.37.0", path = "../gix-index" }
+gix-odb = { version = "^0.66.0", path = "../gix-odb" }
 gix-testtools = { path = "../tests/tools" }

gix-blame/src/error.rs

@@ -0,0 +1,30 @@
+use gix_object::bstr::BString;
+
+/// The error returned by [file()](crate::file()).
+#[derive(Debug, thiserror::Error)]
+#[allow(missing_docs)]
+pub enum Error {
+    #[error("No commit was given")]
+    EmptyTraversal,
+    #[error(transparent)]
+    BlobDiffSetResource(#[from] gix_diff::blob::platform::set_resource::Error),
+    #[error(transparent)]
+    BlobDiffPrepare(#[from] gix_diff::blob::platform::prepare_diff::Error),
+    #[error("The file to blame at '{file_path}' wasn't found in the first commit at {commit_id}")]
+    FileMissing {
+        /// The file-path to the object to blame.
+        file_path: BString,
+        /// The commit whose tree didn't contain `file_path`.
+        commit_id: gix_hash::ObjectId,
+    },
+    #[error("Couldn't find commit or tree in the object database")]
+    FindObject(#[from] gix_object::find::Error),
+    #[error("Could not find existing blob or commit")]
+    FindExistingObject(#[from] gix_object::find::existing_object::Error),
+    #[error("Could not find existing iterator over a tree")]
+    FindExistingIter(#[from] gix_object::find::existing_iter::Error),
+    #[error("Failed to obtain the next commit in the commit-graph traversal")]
+    Traverse(#[source] Box<dyn std::error::Error + Send + Sync>),
+    #[error(transparent)]
+    DiffTree(#[from] gix_diff::tree::Error),
+}