Merge pull request #1585 from Byron/merge

octopus-merge (part 2: blob-merge)

Author: Byron

Cargo.lock

@@ -243,6 +243,7 @@ members = [
     "gix-object",
     "gix-glob",
     "gix-diff",
+    "gix-merge",
     "gix-date",
     "gix-traverse",
     "gix-dir",

Cargo.toml

@@ -130,10 +130,11 @@ is usable to some extent.
   * [gix-submodule](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-submodule)
   * [gix-status](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-status)
   * [gix-worktree-state](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-worktree-state)
-  * `gitoxide-core`
-* **very early**  _(possibly without any documentation and many rough edges)_
   * [gix-date](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-date)
   * [gix-dir](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-dir)
+  * `gitoxide-core`
+* **very early**  _(possibly without any documentation and many rough edges)_
+  * [gix-merge](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-merge)
 * **idea** _(just a name placeholder)_
   * [gix-note](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-note)
   * [gix-fetchhead](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-fetchhead)

README.md

@@ -196,6 +196,9 @@ The top-level crate that acts as hub to all functionality provided by the `gix-*
    * [x] probe capabilities
    * [x] symlink creation and removal
    * [x] file snapshots
+* [ ] **BString Interner with Arena-Backing and arbitrary value association**
+    - probably based on [`internment`](https://docs.rs/internment/latest/internment/struct.Arena.html#),
+      but needs `bumpalo` support to avoid item allocations/boxing, and avoid internal `Mutex`. (key type is pointer based).
 
 ### gix-fs
 * [x] probe capabilities
@@ -215,6 +218,7 @@ The top-level crate that acts as hub to all functionality provided by the `gix-*
       * [x] [name validation][tagname-validation]
 * [x] transform borrowed to owned objects
 * [x] edit trees efficiently and write changes back
+    - [ ] See if `gix-fs::InternedMap` improves performance.
 * [x] API documentation
     * [ ] Some examples
 
@@ -320,11 +324,24 @@ Check out the [performance discussion][gix-diff-performance] as well.
     * [x] prepare invocation of external diff program
         - [ ] pass meta-info
 * [ ] working with hunks of data
+* [ ] diff-heuristics match Git perfectly
 * [x] API documentation
     * [ ] Examples
-    
+
 [gix-diff-performance]: https://github.com/Byron/gitoxide/discussions/74
 
+### gix-merge
+
+* [x] three-way merge analysis of blobs with choice of how to resolve conflicts
+    - [ ] choose how to resolve conflicts on the data-structure
+    - [ ] produce a new blob based on data-structure containing possible resolutions
+        - [x] `merge` style
+        - [x] `diff3` style
+        - [x] `zdiff` style
+* [ ] diff-heuristics match Git perfectly
+* [x] API documentation
+    * [ ] Examples
+
 ### gix-traverse
 
 Check out the [performance discussion][gix-traverse-performance] as well.

crate-status.md

@@ -23,9 +23,9 @@ impl<'a> ValueRef<'a> {
 }
 
 /// Access and conversions
-impl ValueRef<'_> {
+impl<'a> ValueRef<'a> {
     /// Access this value as byte string.
-    pub fn as_bstr(&self) -> &BStr {
+    pub fn as_bstr(&self) -> &'a BStr {
         self.0.as_bytes().as_bstr()
     }
 

gix-attributes/src/state.rs

@@ -22,6 +22,7 @@ pub struct WorktreeRoots {
     pub new_root: Option<PathBuf>,
 }
 
+/// Access
 impl WorktreeRoots {
     /// Return the root path for the given `kind`
     pub fn by_kind(&self, kind: ResourceKind) -> Option<&Path> {
@@ -30,6 +31,11 @@ impl WorktreeRoots {
             ResourceKind::NewOrDestination => self.new_root.as_deref(),
         }
     }
+
+    /// Return `true` if all worktree roots are unset.
+    pub fn is_unset(&self) -> bool {
+        self.new_root.is_none() && self.old_root.is_none()
+    }
 }
 
 /// Data as part of an [Outcome].
@@ -184,6 +190,8 @@ impl Pipeline {
 /// Access
 impl Pipeline {
     /// Return all drivers that this instance was initialized with.
+    ///
+    /// They are sorted by [`name`](Driver::name) to support binary searches.
     pub fn drivers(&self) -> &[super::Driver] {
         &self.drivers
     }
@@ -445,7 +453,7 @@ impl Pipeline {
                                         }
                                     }
                                     .map_err(|err| {
-                                        convert_to_diffable::Error::CreateTempfile {
+                                        convert_to_diffable::Error::StreamCopy {
                                             source: err,
                                             rela_path: rela_path.to_owned(),
                                         }
@@ -533,6 +541,8 @@ impl Driver {
     pub fn prepare_binary_to_text_cmd(&self, path: &Path) -> Option<std::process::Command> {
         let command: &BStr = self.binary_to_text_command.as_ref()?.as_ref();
         let cmd = gix_command::prepare(gix_path::from_bstr(command).into_owned())
+            // TODO: Add support for an actual Context, validate it *can* match Git
+            .with_context(Default::default())
             .with_shell()
             .stdin(Stdio::null())
             .stdout(Stdio::piped())

gix-diff/src/blob/pipeline.rs

@@ -184,7 +184,7 @@ pub mod prepare_diff {
 
     use crate::blob::platform::Resource;
 
-    /// The kind of operation that was performed during the [`diff`](super::Platform::prepare_diff()) operation.
+    /// The kind of operation that should be performed based on the configuration of the resources involved in the diff.
     #[derive(Debug, Copy, Clone, Eq, PartialEq)]
     pub enum Operation<'a> {
         /// The [internal diff algorithm](imara_diff::diff) should be called with the provided arguments.
@@ -383,6 +383,7 @@ impl Platform {
     ///
     /// If one of the resources is binary, the operation reports an error as such resources don't make their data available
     /// which is required for the external diff to run.
+    // TODO: fix this - the diff shouldn't fail if binary (or large) files are used, just copy them into tempfiles.
     pub fn prepare_diff_command(
         &self,
         diff_command: BString,

gix-diff/src/blob/platform.rs

@@ -57,8 +57,10 @@ pub(crate) mod function {
     /// Return `true` if `buf` was written or `false` if nothing had to be done.
     /// Depending on the state in `buf`, `index_object` is called to write the version of `src` as stored in the index
     /// into the buffer and if it is a blob, or return `Ok(None)` if no such object exists.
-    /// If renormalization is desired, let it return `Ok(None)` at all times to not let it have any influence over the
-    /// outcome of this function.
+    ///
+    /// *If renormalization is desired*, let it return `Ok(None)` at all times to not let it have any influence over the
+    /// outcome of this function. Otherwise, it will check if the in-index buffer already has newlines that it would now
+    /// want to change, and avoid doing so as what's in Git should be what's desired (except for when *renormalizing*).
     /// If `round_trip_check` is not `None`, round-tripping will be validated and handled accordingly.
     pub fn convert_to_git(
         src: &[u8],

gix-filter/src/eol/convert_to_git.rs

@@ -91,7 +91,7 @@ impl Pipeline {
             self.options.eol_config,
         )?;
 
-        let mut in_buffer = false;
+        let mut in_src_buffer = false;
         // this is just an approximation, but it's as good as it gets without reading the actual input.
         let would_convert_eol = eol::convert_to_git(
             b"\r\n",
@@ -119,13 +119,13 @@ impl Pipeline {
                 }
                 self.bufs.clear();
                 read.read_to_end(&mut self.bufs.src)?;
-                in_buffer = true;
+                in_src_buffer = true;
             }
         }
-        if !in_buffer && (apply_ident_filter || encoding.is_some() || would_convert_eol) {
+        if !in_src_buffer && (apply_ident_filter || encoding.is_some() || would_convert_eol) {
             self.bufs.clear();
             src.read_to_end(&mut self.bufs.src)?;
-            in_buffer = true;
+            in_src_buffer = true;
         }
 
         if let Some(encoding) = encoding {
@@ -158,7 +158,7 @@ impl Pipeline {
         if apply_ident_filter && ident::undo(&self.bufs.src, &mut self.bufs.dest)? {
             self.bufs.swap();
         }
-        Ok(if in_buffer {
+        Ok(if in_src_buffer {
             ToGitOutcome::Buffer(&self.bufs.src)
         } else {
             ToGitOutcome::Unchanged(src)

gix-filter/src/pipeline/convert.rs

@@ -0,0 +1,49 @@
+[package]
+name = "gix-merge"
+version = "0.0.0"
+repository = "https://github.com/Byron/gitoxide"
+license = "MIT OR Apache-2.0"
+description = "A crate of the gitoxide project implementing merge algorithms"
+authors = ["Sebastian Thiel <[email protected]>"]
+edition = "2021"
+rust-version = "1.65"
+
+[lints]
+workspace = true
+
+[lib]
+doctest = false
+
+[features]
+default = ["blob"]
+## Enable diffing of blobs using imara-diff, which also allows for a generic rewrite tracking implementation.
+blob = ["dep:imara-diff", "dep:gix-filter", "dep:gix-worktree", "dep:gix-path", "dep:gix-fs", "dep:gix-command", "dep:gix-tempfile", "dep:gix-trace", "dep:gix-quote"]
+## Data structures implement `serde::Serialize` and `serde::Deserialize`.
+serde = ["dep:serde", "gix-hash/serde", "gix-object/serde"]
+
+[dependencies]
+gix-hash = { version = "^0.14.2", path = "../gix-hash" }
+gix-object = { version = "^0.44.0", path = "../gix-object" }
+gix-filter = { version = "^0.13.0", path = "../gix-filter", optional = true }
+gix-worktree = { version = "^0.36.0", path = "../gix-worktree", default-features = false, features = ["attributes"], optional = true }
+gix-command = { version = "^0.3.9", path = "../gix-command", optional = true }
+gix-path = { version = "^0.10.11", path = "../gix-path", optional = true }
+gix-fs = { version = "^0.11.3", path = "../gix-fs", optional = true }
+gix-tempfile = { version = "^14.0.0", path = "../gix-tempfile", optional = true }
+gix-trace = { version = "^0.1.10", path = "../gix-trace", optional = true }
+gix-quote = { version = "^0.4.12", path = "../gix-quote", optional = true }
+
+thiserror = "1.0.63"
+imara-diff = { version = "0.1.7", optional = true }
+bstr = { version = "1.5.0", default-features = false }
+serde = { version = "1.0.114", optional = true, default-features = false, features = ["derive"] }
+
+document-features = { version = "0.2.0", optional = true }
+
+[dev-dependencies]
+gix-testtools = { path = "../tests/tools" }
+pretty_assertions = "1.4.0"
+
+[package.metadata.docs.rs]
+all-features = true
+features = ["document-features"]

gix-merge/Cargo.toml

@@ -0,0 +1 @@
+../LICENSE-APACHE

gix-merge/LICENSE-APACHE

@@ -0,0 +1 @@
+../LICENSE-MIT

gix-merge/LICENSE-MIT

@@ -0,0 +1,43 @@
+/// What to do when having to pick a side to resolve a conflict.
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
+pub enum ResolveWith {
+    /// Chose the ancestor to resolve a conflict.
+    Ancestor,
+    /// Chose our side to resolve a conflict.
+    Ours,
+    /// Chose their side to resolve a conflict.
+    Theirs,
+}
+
+/// Tell the caller of [`merge()`](function::merge) which side was picked.
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
+pub enum Pick {
+    /// Chose the ancestor.
+    Ancestor,
+    /// Chose our side.
+    Ours,
+    /// Chose their side.
+    Theirs,
+}
+
+pub(super) mod function {
+    use crate::blob::builtin_driver::binary::{Pick, ResolveWith};
+    use crate::blob::Resolution;
+
+    /// As this algorithm doesn't look at the actual data, it returns a choice solely based on logic.
+    ///
+    /// It always results in a conflict with `current` being picked unless `on_conflict` is not `None`.
+    pub fn merge(on_conflict: Option<ResolveWith>) -> (Pick, Resolution) {
+        match on_conflict {
+            None => (Pick::Ours, Resolution::Conflict),
+            Some(resolve) => (
+                match resolve {
+                    ResolveWith::Ours => Pick::Ours,
+                    ResolveWith::Theirs => Pick::Theirs,
+                    ResolveWith::Ancestor => Pick::Ancestor,
+                },
+                Resolution::Complete,
+            ),
+        }
+    }
+}

gix-merge/src/blob/builtin_driver/binary.rs

@@ -0,0 +1,30 @@
+use crate::blob::BuiltinDriver;
+
+impl BuiltinDriver {
+    /// Return the name of this instance.
+    pub fn as_str(&self) -> &str {
+        match self {
+            BuiltinDriver::Text => "text",
+            BuiltinDriver::Binary => "binary",
+            BuiltinDriver::Union => "union",
+        }
+    }
+
+    /// Get all available built-in drivers.
+    pub fn all() -> &'static [Self] {
+        &[BuiltinDriver::Text, BuiltinDriver::Binary, BuiltinDriver::Union]
+    }
+
+    /// Try to match one of our variants to `name`, case-sensitive, and return its instance.
+    pub fn by_name(name: &str) -> Option<Self> {
+        Self::all().iter().find(|variant| variant.as_str() == name).copied()
+    }
+}
+
+///
+pub mod binary;
+pub use binary::function::merge as binary;
+
+///
+pub mod text;
+pub use text::function::merge as text;