Add feature flag blob-experimental to gix-blame

Author: cruessler

gix-blame/Cargo.toml

@@ -10,12 +10,17 @@ authors = ["Christoph Rüßler <[email protected]>", "Sebastian Thi
 edition = "2021"
 rust-version = "1.82"
 
+[features]
+## An experimental use of the v0.2 branch of `imara-diff` to allow trying it out, and for writing tests against it more easily.
+## We will decide later how it should actually be exposed.
+blob-experimental = ["gix-diff/blob-experimental"]
+
 [dependencies]
 gix-commitgraph = { version = "^0.30.0", path = "../gix-commitgraph" }
 gix-revwalk = { version = "^0.23.0", path = "../gix-revwalk" }
 gix-trace = { version = "^0.1.14", path = "../gix-trace" }
 gix-date = { version = "^0.11.0", path = "../gix-date" }
-gix-diff = { version = "^0.55.0", path = "../gix-diff", default-features = false, features = ["blob", "blob-experimental"] }
+gix-diff = { version = "^0.55.0", path = "../gix-diff", default-features = false, features = ["blob"] }
 gix-object = { version = "^0.52.0", path = "../gix-object" }
 gix-hash = { version = "^0.20.0", path = "../gix-hash" }
 gix-worktree = { version = "^0.44.0", path = "../gix-worktree", default-features = false, features = ["attributes"] }

gix-blame/src/file/function.rs

@@ -1,9 +1,9 @@
 use std::num::NonZeroU32;
 
-use gix_diff::{
-    blob::{intern::TokenSource, v2::Hunk},
-    tree::Visit,
-};
+#[cfg(feature = "blob-experimental")]
+use gix_diff::blob::v2::Hunk;
+
+use gix_diff::{blob::intern::TokenSource, tree::Visit};
 use gix_hash::ObjectId;
 use gix_object::{
     bstr::{BStr, BString},
@@ -751,6 +751,102 @@ fn tree_diff_with_rewrites_at_file_path(
 }
 
 #[allow(clippy::too_many_arguments)]
+#[cfg(not(feature = "blob-experimental"))]
+fn blob_changes(
+    odb: impl gix_object::Find + gix_object::FindHeader,
+    resource_cache: &mut gix_diff::blob::Platform,
+    oid: ObjectId,
+    previous_oid: ObjectId,
+    file_path: &BStr,
+    previous_file_path: &BStr,
+    diff_algorithm: gix_diff::blob::Algorithm,
+    stats: &mut Statistics,
+) -> Result<Vec<Change>, Error> {
+    /// Record all [`Change`]s to learn about additions, deletions and unchanged portions of a *Source File*.
+    struct ChangeRecorder {
+        last_seen_after_end: u32,
+        hunks: Vec<Change>,
+        total_number_of_lines: u32,
+    }
+
+    impl ChangeRecorder {
+        /// `total_number_of_lines` is used to fill in the last unchanged hunk if needed
+        /// so that the entire file is represented by [`Change`].
+        fn new(total_number_of_lines: u32) -> Self {
+            ChangeRecorder {
+                last_seen_after_end: 0,
+                hunks: Vec::new(),
+                total_number_of_lines,
+            }
+        }
+    }
+
+    use std::ops::Range;
+
+    impl gix_diff::blob::Sink for ChangeRecorder {
+        type Out = Vec<Change>;
+
+        fn process_change(&mut self, before: Range<u32>, after: Range<u32>) {
+            // This checks for unchanged hunks.
+            if after.start > self.last_seen_after_end {
+                self.hunks
+                    .push(Change::Unchanged(self.last_seen_after_end..after.start));
+            }
+
+            match (!before.is_empty(), !after.is_empty()) {
+                (_, true) => {
+                    self.hunks.push(Change::AddedOrReplaced(
+                        after.start..after.end,
+                        before.end - before.start,
+                    ));
+                }
+                (true, false) => {
+                    self.hunks.push(Change::Deleted(after.start, before.end - before.start));
+                }
+                (false, false) => unreachable!("BUG: imara-diff provided a non-change"),
+            }
+            self.last_seen_after_end = after.end;
+        }
+
+        fn finish(mut self) -> Self::Out {
+            if self.total_number_of_lines > self.last_seen_after_end {
+                self.hunks
+                    .push(Change::Unchanged(self.last_seen_after_end..self.total_number_of_lines));
+            }
+            self.hunks
+        }
+    }
+
+    resource_cache.set_resource(
+        previous_oid,
+        gix_object::tree::EntryKind::Blob,
+        previous_file_path,
+        gix_diff::blob::ResourceKind::OldOrSource,
+        &odb,
+    )?;
+    resource_cache.set_resource(
+        oid,
+        gix_object::tree::EntryKind::Blob,
+        file_path,
+        gix_diff::blob::ResourceKind::NewOrDestination,
+        &odb,
+    )?;
+
+    let outcome = resource_cache.prepare_diff()?;
+    let input = gix_diff::blob::intern::InternedInput::new(
+        tokens_for_diffing(outcome.old.data.as_slice().unwrap_or_default()),
+        tokens_for_diffing(outcome.new.data.as_slice().unwrap_or_default()),
+    );
+    let number_of_lines_in_destination = input.after.len();
+    let change_recorder = ChangeRecorder::new(number_of_lines_in_destination as u32);
+
+    let res = gix_diff::blob::diff(diff_algorithm, &input, change_recorder);
+    stats.blobs_diffed += 1;
+    Ok(res)
+}
+
+#[allow(clippy::too_many_arguments)]
+#[cfg(feature = "blob-experimental")]
 fn blob_changes(
     odb: impl gix_object::Find + gix_object::FindHeader,
     resource_cache: &mut gix_diff::blob::Platform,

gix-blame/tests/blame.rs

@@ -292,13 +292,27 @@ mktest!(
     3
 );
 
-/// As of 2024-09-24, the Myers-related test was expected to fail. Both tests were initially
-/// written when diffing was done by `imara-diff` 0.1. After updating to `imara-diff` 0.2 on
-/// 2025-12-07, the Myers-related test started passing.
+/// As of 2024-09-24, the Myers-related test is expected to fail. Both tests use `imara-diff` 0.1
+/// under the hood.
 ///
 /// Context: https://github.com/Byron/gitoxide/pull/1453#issuecomment-2371013904
 #[test]
-fn diff_disparity() {
+#[should_panic = "empty-lines-myers"]
+#[cfg(not(feature = "blob-experimental"))]
+fn diff_disparity_imara_diff_v1() {
+    diff_disparity_base()
+}
+
+/// As of 2025-12-07, both tests are expected to pass. They use `imara-diff` 0.2 under the hood.
+///
+/// Context: https://github.com/Byron/gitoxide/pull/1453#issuecomment-2371013904
+#[test]
+#[cfg(feature = "blob-experimental")]
+fn diff_disparity_imara_diff_v2() {
+    diff_disparity_base()
+}
+
+fn diff_disparity_base() {
     for case in ["empty-lines-myers", "empty-lines-histogram"] {
         let Fixture {
             odb,