Support index diffs against the empty tree

git/diff.py

@@ -3,7 +3,7 @@
 # This module is part of GitPython and is released under the
 # 3-Clause BSD License: https://opensource.org/license/bsd-3-clause/
 
-__all__ = ["DiffConstants", "NULL_TREE", "INDEX", "Diffable", "DiffIndex", "Diff"]
+__all__ = ["DiffConstants", "NULL_TREE", "NULL_TREE_SHA", "INDEX", "Diffable", "DiffIndex", "Diff"]
 
 import enum
 import re
@@ -84,6 +84,9 @@ class DiffConstants(enum.Enum):
 :const:`git.NULL_TREE` and :const:`Diffable.NULL_TREE`.
 """
 
+NULL_TREE_SHA = "4b825dc642cb6eb9a060e54bf8d69288fbee4904"
+"""SHA of Git's canonical empty tree object."""
+
 INDEX: Literal[DiffConstants.INDEX] = DiffConstants.INDEX
 """Stand-in indicating you want to diff against the index.
 
@@ -599,7 +602,14 @@ def _index_from_patch_format(cls, repo: "Repo", proc: Union["Popen", "Git.AutoIn
 
  # FIXME: Here SLURPING raw, need to re-phrase header-regexes linewise.
  text_list: List[bytes] = []
- handle_process_output(proc, text_list.append, None, finalize_process, decode_streams=False)
+ stderr_list: List[bytes] = []
+
+ def finalize_process_with_stderr(proc: Union["Popen", "Git.AutoInterrupt"]) -> None:
+ finalize_process(proc, stderr=b"".join(stderr_list))
+
+ handle_process_output(
+ proc, text_list.append, stderr_list.append, finalize_process_with_stderr, decode_streams=False
+ )
 
  # For now, we have to bake the stream.
  text = b"".join(text_list)
@@ -765,11 +775,16 @@ def _index_from_raw_format(cls, repo: "Repo", proc: "Popen") -> "DiffIndex[Diff]
  # :100644 100644 687099101... 37c5e30c8... M .gitignore
 
  index: "DiffIndex" = DiffIndex()
+ stderr_list: List[bytes] = []
+
+ def finalize_process_with_stderr(proc: Union["Popen", "Git.AutoInterrupt"]) -> None:
+ finalize_process(proc, stderr=b"".join(stderr_list))
+
  handle_process_output(
  proc,
  lambda byt: cls._handle_diff_line(byt, repo, index),
-None,
-finalize_process,
+stderr_list.append,
+finalize_process_with_stderr,
  decode_streams=False,
  )
 

git/index/base.py

@@ -1480,12 +1480,11 @@ def reset(
 
  return self
 
- # FIXME: This is documented to accept the same parameters as Diffable.diff, but this
- # does not handle NULL_TREE for `other`. (The suppressed mypy error is about this.)
  def diff(
  self,
- other: Union[ # type: ignore[override]
+ other: Union[
  Literal[git_diff.DiffConstants.INDEX],
+ Literal[git_diff.DiffConstants.NULL_TREE],
  "Tree",
  "Commit",
  str,
@@ -1512,6 +1511,44 @@ def diff(
  if other is self.INDEX:
  return git_diff.DiffIndex()
 
+ if other == git_diff.NULL_TREE or other == git_diff.NULL_TREE_SHA:
+ args: List[Union[PathLike, str]] = [
+ "--cached",
+ git_diff.NULL_TREE_SHA,
+ "--abbrev=40",
+ "--full-index",
+ ]
+
+ if not any(x in kwargs for x in ("find_renames", "no_renames", "M")):
+ args.append("-M")
+
+ if create_patch:
+ args.append("-p")
+ args.append("--no-ext-diff")
+ else:
+ args.append("--raw")
+ args.append("-z")
+
+ args.append("--no-color")
+
+ if paths is not None and not isinstance(paths, (tuple, list)):
+ paths = [paths]
+
+ if paths:
+ args.append("--")
+ args.extend(paths)
+
+ kwargs["as_process"] = True
+ proc = self.repo.git.diff(*args, **kwargs)
+
+ diff_method = (
+ git_diff.Diff._index_from_patch_format if create_patch else git_diff.Diff._index_from_raw_format
+ )
+ index = diff_method(self.repo, proc)
+
+ proc.wait()
+ return index
+
  # Index against anything but None is a reverse diff with the respective item.
  # Handle existing -R flags properly.
  # Transform strings to the object so that we can call diff on it.

test/lib/helper.py

@@ -90,7 +90,7 @@ def __init__(self, input_string):
  self.stdout = io.BytesIO(input_string)
  self.stderr = io.BytesIO()
 
- def wait(self):
+ def wait(self, stderr=None):
  return 0
 
  poll = wait

test/test_index.py

@@ -23,7 +23,8 @@
 import ddt
 import pytest
 
-from git import BlobFilter, Diff, Git, IndexFile, Object, Repo, Tree
+from git import BlobFilter, Diff, Git, IndexFile, NULL_TREE, Object, Repo, Tree
+from git.diff import NULL_TREE_SHA
 from git.exc import (
  CheckoutError,
  GitCommandError,
@@ -555,6 +556,39 @@ def test_index_file_diffing(self, rw_repo):
  rval = index.checkout("lib")
  assert len(list(rval)) > 1
 
+ @with_rw_directory
+ def test_index_file_diff_null_tree_with_initial_index(self, rw_dir):
+ repo = Repo.init(rw_dir)
+ filename = ".gitkeep"
+ file_path = osp.join(repo.working_tree_dir, filename)
+ with open(file_path, "w") as fp:
+ fp.write("# Initial file\n")
+
+ index = repo.index
+ index.add([filename])
+ index.write()
+
+ index = IndexFile(repo)
+ self.assertEqual(len(index.diff(None)), 0)
+
+ diff = index.diff(NULL_TREE)
+ self.assertEqual(len(diff), 1)
+ self.assertEqual(diff[0].change_type, "A")
+ assert diff[0].new_file
+ self.assertEqual(diff[0].b_path, filename)
+
+ self.assertEqual(len(index.diff(NULL_TREE, paths=filename)), 1)
+ self.assertEqual(len(index.diff(NULL_TREE_SHA, paths=filename)), 1)
+ self.assertEqual(len(index.diff(NULL_TREE, paths="missing")), 0)
+
+ patch = index.diff(NULL_TREE, create_patch=True)
+ self.assertEqual(len(patch), 1)
+ self.assertIn(b"+# Initial file", patch[0].diff)
+
+ with self.assertRaises(GitCommandError) as exc_info:
+ index.diff(NULL_TREE, bogus_option=True)
+ self.assertIn("usage: git diff", exc_info.exception.stderr)
+
  def _count_existing(self, repo, files):
  """Return count of files that actually exist in the repository directory."""
  existing = 0

test/test_remote.py

@@ -687,7 +687,12 @@ def test_multiple_urls(self, rw_repo):
 
  def test_fetch_error(self):
  rem = self.rorepo.remote("origin")
- with self.assertRaisesRegex(GitCommandError, "[Cc]ouldn't find remote ref __BAD_REF__"):
+ msg = (
+ r"[Cc]ouldn't find remote ref __BAD_REF__|"
+ r"could not read Username|"
+ r"expected flush after ref listing"
+ )
+ with self.assertRaisesRegex(GitCommandError, msg):
  rem.fetch("__BAD_REF__")
 
  @with_rw_repo("0.1.6", bare=False)