Merge pull request GitoxideLabs#2420 from cruessler/remove-imara-diff-0-1-in-gix-blame

Switch to `imara-diff` 0.2 in `gix-blame`

Author: Byron

Cargo.toml

@@ -155,11 +155,6 @@ http-client-reqwest = ["gix/blocking-http-transport-reqwest-rust-tls"]
 ## Use async client networking.
 gitoxide-core-async-client = ["gitoxide-core/async-client", "futures-lite"]
 
-#! ### Experimental Features
-#!
-## An experimental use of the v0.2 branch of `imara-diff` in `gix-diff` and `gix-blame`, to allow trying the new diff algorithm.
-blame-experimental = ["gitoxide-core/blame-experimental"]
-
 [dependencies]
 anyhow = "1.0.98"
 

gitoxide-core/Cargo.toml

@@ -46,10 +46,6 @@ async-client = ["gix/async-network-client-async-std", "gix-transport-configurati
 ## Data structures implement `serde::Serialize` and `serde::Deserialize`.
 serde = ["gix/serde", "dep:serde_json", "dep:serde", "bytesize/serde"]
 
-#! ### Experimental
-## An experimental use of the v0.2 branch of `imara-diff` in `gix-diff` and `gix-blame`, to allow trying the new diff algorithm.
-blame-experimental = ["gix/blame-experimental"]
-
 [dependencies]
 # deselect everything else (like "performance") as this should be controllable by the parent application.
 gix = { version = "^0.78.0", path = "../gix", default-features = false, features = ["merge", "blob-diff", "blame", "revision", "mailmap", "excludes", "attributes", "worktree-mutation", "credentials", "interrupt", "status", "dirwalk"] }

gix-blame/Cargo.toml

@@ -10,18 +10,13 @@ authors = ["Christoph Rüßler <christoph.ruessler@mailbox.org>", "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-error = { version = "^0.0.0", path = "../gix-error" }
 gix-commitgraph = { version = "^0.32.0", path = "../gix-commitgraph" }
 gix-revwalk = { version = "^0.26.0", path = "../gix-revwalk" }
 gix-trace = { version = "^0.1.17", path = "../gix-trace" }
 gix-date = { version = "^0.13.0", path = "../gix-date" }
-gix-diff = { version = "^0.58.0", path = "../gix-diff", default-features = false, features = ["blob"] }
+gix-diff = { version = "^0.58.0", path = "../gix-diff", default-features = false, features = ["blob", "blob-experimental"] }
 gix-object = { version = "^0.55.0", path = "../gix-object" }
 gix-hash = { version = "^0.22.0", path = "../gix-hash" }
 gix-worktree = { version = "^0.47.0", path = "../gix-worktree", default-features = false, features = ["attributes"] }

gix-blame/src/file/function.rs

@@ -748,102 +748,6 @@ 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> {
-    use std::ops::Range;
-
-    /// 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,
-            }
-        }
-    }
-
-    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

@@ -307,28 +307,16 @@ mktest!(
     3
 );
 
-/// As of 2024-09-24, the Myers-related test is expected to fail. Both tests use `imara-diff` 0.1
-/// under the hood.
+/// As of 2025-12-07, both algorithms are expected to pass. They use `imara-diff` 0.2 under the
+/// hood. One of them failed with `imara-diff` 0.1.
 ///
 /// Context: https://github.com/Byron/gitoxide/pull/1453#issuecomment-2371013904
 #[test]
-#[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 algorithms 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"] {
+fn diff_algorithm_parity() {
+    for (case, diff_algorithm) in [
+        ("empty-lines-myers", gix_diff::blob::Algorithm::Myers),
+        ("empty-lines-histogram", gix_diff::blob::Algorithm::Histogram),
+    ] {
         let Fixture {
             odb,
             mut resource_cache,
@@ -344,7 +332,7 @@ fn diff_disparity_base() {
             &mut resource_cache,
             source_file_name.as_ref(),
             gix_blame::Options {
-                diff_algorithm: gix_diff::blob::Algorithm::Histogram,
+                diff_algorithm,
                 ranges: BlameRanges::default(),
                 since: None,
                 rewrites: Some(gix_diff::Rewrites::default()),

gix/Cargo.toml

@@ -318,11 +318,6 @@ progress-tree = ["prodash/progress-tree"]
 ## Print debugging information about usage of object database caches, useful for tuning cache sizes.
 cache-efficiency-debug = ["gix-features/cache-efficiency-debug"]
 
-#! #### Experimental Features
-## Use the v0.2 branch of `imara-diff` in `gix-diff` and `gix-blame`, to allow trying the new diff algorithm.
-blame-experimental = ["blame", "gix-blame/blob-experimental"]
-
-
 [dependencies]
 gix-error = { version = "^0.0.0", path = "../gix-error" }
 gix-utils = { version = "^0.3.1", path = "../gix-utils" }

justfile

@@ -180,7 +180,7 @@ unit-tests:
     cargo nextest run -p gix-transport --features async-client --no-fail-fast
     cargo nextest run -p gix-protocol --features blocking-client --no-fail-fast
     cargo nextest run -p gix-protocol --features async-client --no-fail-fast
-    cargo nextest run -p gix-blame --features blob-experimental --no-fail-fast
+    cargo nextest run -p gix-blame --no-fail-fast
     cargo nextest run -p gix --no-default-features --no-fail-fast
     cargo nextest run -p gix --no-default-features --features basic,comfort,max-performance-safe --no-fail-fast
     cargo nextest run -p gix --no-default-features --features basic,extras,comfort,need-more-recent-msrv --no-fail-fast