How to Fix an Oversized Git Repository Caused by Commit History

I recently needed to push an open-source project to the community, but the repository was reported as too large, several hundred MB. After downloading the current files, I found they were only about 10 MB, and there were only two branches. So I suspected the repository size problem came from Git history.

Install

brew install git-filter-repo

Find Which Paths Are Large

git rev-list --objects --all \
| git cat-file --batch-check='%(objectname) %(objecttype) %(objectsize) %(rest)' \
| grep ' blob ' \
| sort -k3 -nr \
| head -20 \
| awk '{
  size=$3;
  path=$4;
  for (i=5; i<=NF; i++) path=path " " $i;
  if (size >= 1024*1024*1024)
    printf "%.2f GB\t%s\n", size/1024/1024/1024, path;
  else
    printf "%.2f MB\t%s\n", size/1024/1024, path;
}'

Solution

1. git filter-repo –path xxx –invert-paths –force
2. git push –force

Conclusion

1. A large repo does not mean the current files are large.
2. Large files may still exist in history, even if you have already deleted them.