misc.add_argument('--uuid-erase', metavar='UUID',
help=_("Completely erase all traces of a specific UUID from history, including commits"))
misc.add_argument('--notebook-erase', metavar='NOTEBOOK_UUID:UUID1,UUID2,...',
help=_("Completely erase an entire notebook and all its UUIDs from history"))
last part of git-filter-repo
class UUIDEraseFilter(RepoFilter):
"""
Specialized filter for completely erasing UUIDs and their associated commits
"""
def __init__(self, args, uuid_to_erase):
super().__init__(args)
self.uuid = uuid_to_erase.encode() if isinstance(uuid_to_erase, str) else uuid_to_erase
self.commits_removed = 0
self.blobs_removed = 0
def _tweak_commit(self, commit, aux_info):
"""Override to aggressively remove UUID commits"""
# CRITICAL: Check if this commit contains the UUID in its message
if commit.message and self.uuid in commit.message:
# Skip ERASED commits (keep tombstones)
if b"ERASED" not in commit.message:
# 🔇 Verbose output - comment out for silent operation
# print(f" 🔥 REMOVING commit {commit.original_id[:8] if commit.original_id else 'unknown'} (contains UUID)")
commit.skip() # This removes the entire commit
self.commits_removed += 1
return
# ========== FIX: Handle missing parents gracefully ==========
# Filter out parents that are not in the graph
orig_parents = aux_info.get('orig_parents', [])
valid_parents = []
for p in orig_parents:
if p in self._orig_graph.value:
valid_parents.append(p)
if len(valid_parents) != len(orig_parents):
# Some parents missing, update the commit's parents
commit.parents = valid_parents
aux_info['orig_parents'] = valid_parents
# ========== END FIX ==========
# Also check file changes for the UUID
has_uuid_in_files = False
for change in commit.file_changes:
if change.type == b'M' and hasattr(change, 'blob_id'):
# We'll let the blob_callback handle content removal
pass
# Call parent method to handle normal filtering
try:
super()._tweak_commit(commit, aux_info)
except AssertionError:
# If assertion fails, skip this commit
commit.skip()
self.commits_removed += 1
def _tweak_blob(self, blob):
"""Remove UUID from blob contents"""
if blob.data and self.uuid in blob.data:
# Count how many blobs we're modifying
if self.uuid in blob.data:
old_size = len(blob.data)
blob.data = blob.data.replace(self.uuid, b'')
new_size = len(blob.data)
if new_size < old_size:
self.blobs_removed += 1
# 🔇 Verbose output - comment out for silent operation
# print(f" 📦 Removed UUID from blob {blob.original_id[:8] if blob.original_id else 'unknown'}")
# Call parent method
super()._tweak_blob(blob)
class NotebookEraseFilter(RepoFilter):
"""
Specialized filter for completely erasing an entire notebook and all its commits
"""
def __init__(self, args, notebook_uuid, all_uuids):
super().__init__(args)
self.notebook_uuid = notebook_uuid.encode() if isinstance(notebook_uuid, str) else notebook_uuid
self.all_uuids = [u.encode() if isinstance(u, str) else u for u in all_uuids]
self.commits_removed = 0
def _tweak_commit(self, commit, aux_info):
"""Override to aggressively remove ALL commits belonging to this notebook"""
# Get the full commit message
full_message = commit.message or b''
# Check if this commit belongs to the notebook by looking for root UUID
root_pattern = b'root:' + self.notebook_uuid
if root_pattern in full_message:
commit.skip()
self.commits_removed += 1
return
# Also check for any UUID from this notebook
for uuid in self.all_uuids:
uuid_pattern = b'uuid:' + uuid
if uuid_pattern in full_message:
commit.skip()
self.commits_removed += 1
return
# ========== FIX: Filter out parents that are not in the graph ==========
orig_parents = aux_info.get('orig_parents', [])
valid_parents = []
for p in orig_parents:
# Check if this parent exists in the graph
if hasattr(self, '_orig_graph') and hasattr(self._orig_graph, 'value'):
if p in self._orig_graph.value:
valid_parents.append(p)
else:
# Parent was filtered out - this is expected during erasure
pass
else:
# Fallback: keep parent if we can't check
valid_parents.append(p)
# Update the commit's parents to only valid ones
if len(valid_parents) != len(orig_parents):
commit.parents = valid_parents
aux_info['orig_parents'] = valid_parents
# If no valid parents remain, this commit becomes a root commit
# That's fine - it will still be processed
# ========== END FIX ==========
# Call parent for normal processing
try:
super()._tweak_commit(commit, aux_info)
except AssertionError as e:
# If assertion still fails, skip this commit
commit.skip()
self.commits_removed += 1
def main():
setup_gettext()
args = FilteringOptions.parse_args(sys.argv[1:])
# Check for notebook erasure mode
if hasattr(args, 'notebook_erase') and args.notebook_erase:
# Format: notebook_erase=notebook_uuid:uuid1,uuid2,uuid3
parts = args.notebook_erase.split(':')
notebook_uuid = parts[0]
all_uuids = parts[1].split(',') if len(parts) > 1 else [notebook_uuid]
filter = NotebookEraseFilter(args, notebook_uuid, all_uuids)
filter.run()
print(f"\n ✓ Notebook completely erased")
print(f" Removed {filter.commits_removed} commit(s)")
return
# Check for UUID erasure mode (single item)
if hasattr(args, 'uuid_erase') and args.uuid_erase:
filter = UUIDEraseFilter(args, args.uuid_erase)
filter.run()
print(f"\n ✓ UUID {args.uuid_erase} completely erased from history")
print(f" Removed {filter.commits_removed} commit(s)")
return
if args.analyze:
RepoAnalyze.run(args)
else:
filter = RepoFilter(args)
filter.run()
if __name__ == '__main__':
main()
else:
# Allow importing as a module
pass
from cli, (i never used it outside where it is embedded and used as python module)
python git_filter_repo.py --uuid-erase <UUID> --force
Example:
python git_filter_repo.py --uuid-erase 20260609174733 --force
or
python git_filter_repo.py --notebook-erase <NOTEBOOK_UUID>:<UUID1>,<UUID2>,... --force
python git_filter_repo.py --notebook-erase 20262609175249:20260609174733,20262609175249,20260609174738 --force
i believe git filter repo can be used safely for many such problem like this. i solve the parmanent erasing issue using git-filter-repo in my app. I did this modification directing llm. The main requirement is to have uuid in commit structure consistantly. I am no python or git internal expert. I am a sysadmin with 25 years of system knowledge. and i noticed that it is also gdpr compatible and can be used elsewhere based on this concept, what do you think for the future usages in other environment?