Git can do JSON (almost)
If you wish to do some processing of git logs using JSON as an intermediate format, you can add this to the [alias] section of you ~/.gitconfig:
json = log --format='{ \"hashes\":{ \"commit\":\"%H\", \"tree\":\"%T\", \"parents\":\"%P\" }, \"author\":{ \"date\": \"%ai\", \"name\": \"%an\", \"email\":\"%ae\" }, \"committer\":{ \"date\": \"%ci\", \"name\": \"%cn\", \"email\":\"%ce\" } }'This will give you an object-per-line without commas, it will not contain the commit message body or subject, since that is's not mean for humans to read (if you need it, add \"message\":\"%B\" at your own risk of dealing with \ns).
To get the JSON output serialised into an array, do this:
git --no-pager json | awk \
'BEGIN { print("[") } \
{ print($0",") } \
END { print("\"undefined\"]") }'The man page doesn't say whether there is or there isn't a special character placeholder character for getting the metris of LOC or files being added/deleted, though there must be some hacky way of getting it out. Let me know if you do find this one out!
Written by Ilya Dmitrichenko
Related protips
4 Responses
That was a good take!
The most straight-forward way to output stats (through --shortstat) is cumbersome as it gets printed on a new line. As some commits don't have any stats, the new line pattern breaks, making it difficult to port the data reliably. To compensate for that, I had to use several bash commands besides the git log (sed, tr, paste), as you also realised was necessary to use awk...
My solution involves two scripts and is a bit too lengthy to post here, so I'm posting a link instead:
https://github.com/dreamyguy/gitlogg
Some of Gitlogg's features are:
- Parse the
git logof multiple repositories into oneJSONfile. - Introduced
repositorykey/value. - Introduced
files changed,insertionsanddeletionskeys/values. - Introduced
impactkey/value, which represents the cumulative changes for the commit (insertions-deletions). - Sanitise double quotes
"by converting them to single quotes'on all values that allow or are created by user input, likesubject. - Nearly all the
pretty=format:placeholders are available. - Easily include / exclude which keys/values will be parsed to
JSONby commenting out/uncommenting the available ones. - Easy to read code that's thoroughly commented.
- Script execution feedback on console.
- Error handling (since path to repositories needs to be set correctly).
You can use \"subject\":\"%s\" for the first line of the commit message and not have to handle newlines.
JC now has a git log parser that supports various pretty formats and --stat and --shortstat options.
