Week 2 - M2: Real index file + working status

[aman-a-shah]

Owner: @Saanvi521
Module: 2 - Index & working tree
File you own: minigit/index.py
Week: 2 of 10 - Real Bytes on Disk

What changes this week

• Week 1 index lived in self._entries and vanished when the process exited
• Week 2 the index is a real file: .minigit/index
• minigit add then minigit status in two separate commands must now work
• @mahis1067's real ObjectStore lands this week - use it, stop faking blobs

Index file format - exact, do not change alone

<mode> <blob-hash> <path>\n

• one line per entry, sorted by path, UTF-8, trailing \n on every line
• paths are repo-relative and use / (src/utils/parse.py), never absolute

Steps

1. Branch week2/m2-real-index (branch off main after M1 merges)
2. read_index -> index file missing = return []; else parse each line with line.split(" ", 2) -> IndexEntry, return sorted by path
   • split(" ", 2) and not plain split() - filenames can contain spaces
3. write_index -> sort by path, join lines, mkdir(parents=True, exist_ok=True), write the file
   • drop self._entries entirely; the file is now the only state
4. stage_file(path):
   • normalise to a repo-relative path, missing file -> MiniGitError
   • read bytes -> self.store.write_object(data, "blob")
   • read_index, drop any entry with the same path, append IndexEntry("100644", blob_hash, path), write_index
   • executable bit set on disk -> mode "100755" (os.access(p, os.X_OK))
   • snapshot rule still holds: editing the file after staging does not change the index
5. diff_working_tree_vs stays a stub, but make it real against the index for status:
   • add _working_status() -> DiffResult - for each index entry, re-hash the file on disk
     • file gone -> deleted
     • hash differs -> modified
     • files on disk not in the index and not under .minigit/ -> added (untracked)
6. minigit status prints, in this order, one path per line:

staged:
 <path>
not staged:
 <path>
untracked:
 <path>

• nothing in a section -> skip the whole section; nothing at all -> print clean

7. build_tree_from_index and checkout stay stubs # Week 3 / Week 4
8. Tests tests/test_index.py - delete FakeObjectStore, use the real ObjectStore(tmp_path):
   • stage a file, new WorkingTree on the same path, read_index still has it
   • staging twice replaces, does not duplicate
   • index file on disk has exactly one line, sorted order across two files
   • path with a space round-trips
   • edit after staging -> shows up as not staged, index hash unchanged
   • untracked file appears in untracked
   • deleting a staged file -> deleted
9. quality-check green -> commit, push, PR

Notes

• run minigit init first in tests (tmp_path), or create .minigit/ yourself
• @DanisLol calls build_tree_from_index() and checkout() - names stay exact

Done when

• index is a real file, format above
• state survives across two CLI calls
• real ObjectStore, no fake
• status shows staged / not staged / untracked
• tests cover disk round-trip + spaces in paths
• quality-check green
• PR open + reviewed