Week 2 - M1: Real SHA-1 + zlib object store on disk

[aman-a-shah]

Owner: @mahis1067
Module: 1 - Object storage
File you own: minigit/objects.py
Week: 2 of 10 - Real Bytes on Disk

What changes this week

• Week 1 was fake: crc32 hash, dict in memory
• Week 2 is real: SHA-1, zlib, files under .minigit/objects/
• signatures do not change - everyone else's code keeps working

Steps

1. Branch: git checkout main && git pull && git checkout -b week2/m1-real-object-store
2. Delete self._fake_store and the crc32 placeholder
3. hash_object -> hashlib.sha1(f"{obj_type} {len(data)}".encode() + b"\0" + data).hexdigest()
   • the header is inside the hash - blob 2\0hi, not just hi
4. _object_path(hash) -> self.objects_dir / hash[:2] / hash[2:]
   • 2-char folder split keeps one directory from holding 100k files
5. write_object -> hash, mkdir(parents=True, exist_ok=True), zlib.compress(header + data), write bytes
   • path already exists -> return the hash, write nothing (same bytes = same object)
6. read_object -> missing path = ObjectNotFoundError(hash)
   • read, zlib.decompress, split once on b"\0" -> header + content
   • parse "<type> <len>" from the header
   • re-hash the content, mismatch -> ObjectCorruptError
   • return (type, content)
7. New command minigit init (you own it, @DanisLol owns what goes in HEAD):
   • create .minigit/, .minigit/objects/, .minigit/refs/heads/, empty .minigit/index, empty .minigit/config
   • write .minigit/HEAD = ref: refs/heads/main\n
   • already a repo -> print already a minigit repository and return 0, do not wipe anything
8. Drop the module-level _cli_store singleton - build an ObjectStore(".") inside each handler instead
9. Tests - all on tmp_path, no fakes anymore:
   • hash_object(b"hi", "blob") == "32f95c0d1244a78b2be1bab8de17906fabb2c4a8"
   • hash_object(b"", "blob") == "e69de29bb2d1d6434b8b29ae775ad8c2e48c5391"
   • round-trip through disk returns the same bytes
   • the file on disk is compressed (raw bytes != content)
   • writing twice does not error and leaves one file
   • unknown hash -> ObjectNotFoundError
   • overwrite an object file with junk -> ObjectCorruptError
   • init creates every path, running it twice is safe
10. scripts/quality-check.sh green -> commit, push, PR into main

Notes

• those two hashes match real git hash-object output - check with printf hi | git hash-object --stdin
• you are still the base of the stack - land first

Done when

• SHA-1 with the <type> <len>\0 header
• objects zlib-compressed at objects/ab/cdef...
• corrupt object detected on read
• minigit init scaffolds a repo, idempotent
• tests use real disk, no dict
• quality-check green
• PR open + reviewed