Fantaisies programatico-ludiques

Retrieving a file version (git commit & tag) based on its hash

Today, I did some tests on a server where an old version of our project was deployed. At some point, I needed to identify which version of the code was there, and I wrote a pretty shell function to figure this out.

Yeah, I know what your thinking : there must be another way. Surely this information should be somewhere in your deployment system. Probably. But that's no excuse to skip a good git scripting exercise !

Ok, so here is the situation :

  • a git repository cloned on my development machine
  • a remote server without git installed
  • a readable file on this server that originates from the git repository

Now, the idea is to compute the file hash similarly to git, and then find a commit for which git rev-parse returns the same hash.

First thing first, lets build a function that does the same as git hash-object but does not require git. You can check this SO answer or the details on how git compute blob hashes. Here is my solution:

git-hash-object () { # USAGE: git-hash-object [-t type] file
    local type=blob
    [ "$1" = "-t" ] && shift && type=$1 && shift
    # depending on your git eol/autocrlf settings,
    # you may want to substitute CRLFs by LFs first,
    # by using `perl -pe 's/\r$//g'` instead of `cat` in the next 2 commands
    local size=$(cat $1 | wc -c | sed 's/ .*$//')
    ( echo -en "$type $size\0"; cat "$1" ) | sha1sum | sed 's/ .*$//'

Ok, lets write our search function now:

git-identify-filehash () { # USAGE: git-identify file hash
    local file=${1?}
    local hash=${2?}
    git log --format="%h %s" $file | while read commit msg; do
        if [ $(git rev-parse $commit:$file) = "$hash" ]; then
            echo $commit $msg - oldest tag including this commit: \
                $(git tag --contains $commit | head -n 1)

Ok, I'm going to try this on the second-to-last version of pre-commit, that is from commit c3c98af at the time I'm writing :

$ wget
$ git hash-object
# Lets pretend we are on a remote server without git:
$ git-hash-object

# Now lets identify this file hash in our repository clone
$ git clone
$ git-identify-filehash pre_commit/ e292c72c6c86e8809bd792a630e7f90ac811c385
c3c98af Support pre-commit from inside submodules - oldest tag including this commit: v0.7.1

Oh My God