removing files from git history that have changed name or path

Question

I am trying to identify large files in the history of my project that have been deleted. When I do this I can see an ordered list.

$ git rev-list --objects --all \
  | git cat-file --batch-check='%(objecttype) %(objectname) %(objectsize) %(rest)' \
  | sed -n 's/^blob //p' \
  | sort --numeric-sort --key=2 \
  | cut -c 42-

Usually, I can remove a file from history without issue like this:

git filter-repo --force --invert-paths --path <path-to-file>

However, it would appear that if a file had a different name or path in the past, once I delete the file, when I run the rev-list again I see the same file I just deleted, with the same size, but with a different path or name it had in the past.

Is there a way to see all file paths of objects with rev-list so that I don't have to iterate with path/name changes in the history like this?

Answer 1

A not very satisfying way would be to run git ls-tree -r on all commits in your repo, and grep the hash of the blob you are looking for :

git rev-list --all | xargs -L1 git ls-tree -r | grep "$hash"
# you can also replace 'xargs' with 'parallel'

# you can insert a command to avoid noisy repetitions of the same path :
... | awk '{ if (!seen[$4]++) { print $4 } }'

---

The answer to this question could help :

• Which commit has this blob?

If you replace the 0|1 result of the check_tree function with the list of paths at which the blob is found within the tree, the memoization of the function should also work just fine.

---

Note that the git filter-repo command you run removes any version of the file located at <path-to-file> from your history, not just the ones when that file's content matches exactly the blob you identified.