Write V2 Index Files (with TREE and EOIE extensions)

[SidneyDouw]

Write V2 Index Files

• Link to git Documentation
• git implementation

Features

Version Support

• V2

Extensions

• tree cache
• end of index entry

Out of Scope

• V3
• V4
• resolve undo
• split index / link
• untracked cache
• fs monitor
• index entry offset table

[Byron]

Session Notes

• File::write_to() to write with trailing hash
• outscope everything that you didn't start yet
• refactor tests roundtrip
• pass std::io::Write to write_to as first argument, see other write_to() methods for reference
• write docs about Options::hash_kind - it's kind of intricate
• could there be baseline tests against all existing fixtures somewhat like this?
• definitely get rid of the index-table, you could try to wrap std::io::Write like so to track bytes written for example.

diff --git a/git-index/tests/index/file/write.rs b/git-index/tests/index/file/write.rs
index aa4342816..49d2129a1 100644
--- a/git-index/tests/index/file/write.rs
+++ b/git-index/tests/index/file/write.rs
@@ -1,5 +1,17 @@
 use git_index::write::Options;
 
+#[test]
+fn roundtrips() {
+    for version in [Version::V2] {
+        for fixture_path in ["all fixtures"] {
+            let mut buf = Vec::<u8>::new();
+            let actual = git_index::File::at(fixture_path)?;
+            actual.state.write_to(&mut buf, Options { version })?;
+            let foo = git_index::State::from_bytes(&buf)?;
+        }
+    }
+}
+
 #[test]
 fn v2_empty() {
     let path = crate::fixture_index_path("V2_empty");

[Byron]

Session Notes

• next step would be creating an index from a tree, as it's something needed on the critical path towards supporting clones.

[Byron]

Thanks a lot!

I took a couple of notes in the tracking issue and plan to look at it a little more thoroughly to generalize the end-of-index extension a little more.

The main refactorings done here should help to…

• …put extension write implementations to the extension itself
• …add documentation to new bits and pieces as allow(missing_docs) was now removed. Doing that helped to recap what's actually present in this crate as I haven't touched it in a while.

What's currently in main should be a good basis for the read-tree work.