Merge pull request #1834 from GitoxideLabs/improvements

improvements

Author: Byron

gix/src/filter.rs

@@ -4,6 +4,7 @@ use std::borrow::Cow;
 pub use gix_filter as plumbing;
 use gix_object::Find;
 
+use crate::prelude::ObjectIdExt;
 use crate::{
     bstr::BStr,
     config::{
@@ -207,6 +208,9 @@ impl Pipeline<'_> {
     }
 
     /// Add the worktree file at `rela_path` to the object database and return its `(id, entry, symlink_metadata)` for use in a tree or in the index, for instance.
+    /// If `rela_path` is a directory *and* is a repository with its `HEAD` pointing to a commit, it will also be provided with the appropriate kind.
+    /// Note that this can easily lead to embedded repositories as no submodule-crosscheck is performed. Otherwise, unreadable repositories or directories
+    /// are ignored with `None` as return value.
     ///
     /// `index` is used in particularly rare cases where the CRLF filter in auto-mode tries to determine whether to apply itself,
     /// and it should match the state used when [instantiating this instance](Self::new()).
@@ -257,6 +261,14 @@ impl Pipeline<'_> {
                 gix_object::tree::EntryKind::Blob
             };
             (id, kind)
+        } else if md.is_dir() {
+            let Some(submodule_repo) = crate::open_opts(&path, repo.open_options().clone()).ok() else {
+                return Ok(None);
+            };
+            let Some(id) = submodule_repo.head_id().ok() else {
+                return Ok(None);
+            };
+            (id.detach().attach(repo), gix_object::tree::EntryKind::Commit)
         } else {
             // This is probably a type-change to something we can't track.
             return Ok(None);

gix/tests/fixtures/repo_with_untracked_files.sh

@@ -1,9 +1,19 @@
 #!/usr/bin/env bash
 set -eu -o pipefail
 
+git init embedded-repository
+(cd embedded-repository
+  echo content >file && git add file && git commit -m "init"
+)
+
 git init -q
 echo content >file
 ln -s file link
 
 echo binary >exe && chmod +x exe
 mkfifo fifo
+
+git submodule add ./embedded-repository submodule
+
+mkdir empty-dir
+git init uninitialized-embedded-repository

gix/tests/gix/repository/filter.rs

@@ -35,11 +35,32 @@ fn pipeline_in_repo_without_special_options() -> crate::Result {
 #[cfg(unix)]
 fn pipeline_worktree_file_to_object() -> crate::Result {
     let repo = named_repo("repo_with_untracked_files.sh")?;
+    let work_dir = repo.work_dir().expect("non-bare");
     let (mut pipe, index) = repo.filter_pipeline(None)?;
     fn take_two<A, B, C>(t: Option<(A, B, C)>) -> Option<(A, B)> {
         t.map(|t| (t.0, t.1))
     }
 
+    let submodule_id = hex_to_id("a047f8183ba2bb7eb00ef89e60050c5fde740483");
+    assert_eq!(
+        take_two(pipe.worktree_file_to_object("embedded-repository".into(), &index)?),
+        Some((submodule_id, gix::object::tree::EntryKind::Commit))
+    );
+    assert_eq!(
+        take_two(pipe.worktree_file_to_object("submodule".into(), &index)?),
+        Some((submodule_id, gix::object::tree::EntryKind::Commit))
+    );
+    assert_eq!(
+        take_two(pipe.worktree_file_to_object("uninitialized-embedded-repository".into(), &index)?),
+        None,
+        "repositories that don't have HEAD pointing to an ID yet are ignored"
+    );
+    assert!(work_dir.join("empty-dir").is_dir());
+    assert_eq!(
+        take_two(pipe.worktree_file_to_object("empty-dir".into(), &index)?),
+        None,
+        "directories that aren't even repos are also ignored"
+    );
     assert_eq!(
         take_two(pipe.worktree_file_to_object("file".into(), &index)?),
         Some((
@@ -66,10 +87,7 @@ fn pipeline_worktree_file_to_object() -> crate::Result {
         None,
         "Missing files are specifically typed and no error"
     );
-    assert!(
-        repo.work_dir().expect("non-bare").join("fifo").exists(),
-        "there is a fifo"
-    );
+    assert!(work_dir.join("fifo").exists(), "there is a fifo");
     assert_eq!(
         take_two(pipe.worktree_file_to_object("fifo".into(), &index)?),
         None,