Pretty-print git repository collaborators sorted by contributions.
~$ git fame --cost hour,month --loc ins
Processing: 100%|โโโโโโโโโโโโโโโโโโโโโโโโโโ| 1/1 [00:00<00:00, 2.16repo/s]
Total commits: 1775
Total ctimes: 2770
Total files: 461
Total hours: 449.7
Total loc: 41659
Total months: 151.0
| Author | hrs | mths | loc | coms | fils | distribution |
|:---------------------|------:|-------:|------:|-------:|-------:|:----------------|
| Casper da Costa-Luis | 228 | 108 | 28572 | 1314 | 172 | 68.6/74.0/37.3 |
| Stephen Larroque | 28 | 18 | 5243 | 203 | 25 | 12.6/11.4/ 5.4 |
| pgajdos | 2 | 9 | 2606 | 2 | 18 | 6.3/ 0.1/ 3.9 |
| Martin Zugnoni | 2 | 5 | 1656 | 3 | 3 | 4.0/ 0.2/ 0.7 |
| Kyle Altendorf | 7 | 2 | 541 | 31 | 7 | 1.3/ 1.7/ 1.5 |
| Hadrien Mary | 5 | 1 | 469 | 31 | 17 | 1.1/ 1.7/ 3.7 |
| Richard Sheridan | 2 | 1 | 437 | 23 | 3 | 1.0/ 1.3/ 0.7 |
| Guangshuo Chen | 3 | 1 | 321 | 18 | 7 | 0.8/ 1.0/ 1.5 |
| Noam Yorav-Raphael | 4 | 1 | 229 | 11 | 6 | 0.5/ 0.6/ 1.3 |
| github-actions[bot] | 2 | 1 | 186 | 1 | 51 | 0.4/ 0.1/11.1 |
...The distribution column is a percentage breakdown of loc/coms/fils. (e.g. in the table above, Casper has written surviving code in 172/461 = 37.3% of all files).
Table of contents
Pull and install:
On Windows, run:
This is probably not necessary on UNIX systems. If git fame doesn't work after restarting the terminal on Linux & Mac OS, try (with single quotes):
Optionally, systems with bash-completion can install tab completion support. The git-fame_completion.bash file needs to be copied to an appropriate folder.
On Ubuntu, the procedure would be:
followed by a terminal restart.
The list of all changes is available on the Releases page:
For example, to print statistics regarding all source files in a C++/CUDA repository (*.c/h/t(pp), *.cu(h)), carefully handling whitespace and line copies:
It is also possible to run from within a python shell or script.
Usage:
git-fame [--help | options] [<gitdir>...]
Arguments:
<gitdir> Git directory [default: ./].
May be specified multiple times to aggregate across
multiple repositories.
Options:
-h, --help Print this help and exit.
-v, --version Print module version and exit.
--branch=<b> Branch or tag [default: HEAD] up to which to check.
--sort=<key> [default: loc]|commits|files|hours|months.
--loc=<type> surv(iving)|ins(ertions)|del(etions)
What `loc` represents. Use 'ins,del' to count both.
defaults to 'surviving' unless `--cost` is specified.
--excl=<f> Excluded files (default: None).
In no-regex mode, may be a comma-separated list.
Escape (\,) for a literal comma (may require \\, in shell).
--incl=<f> Included files [default: .*]. See `--excl` for format.
--since=<date> Date from which to check. Can be absoulte (eg: 1970-01-31)
or relative to now (eg: 3.weeks).
--cost=<method> Include time cost in person-months (COCOMO) or
person-hours (based on commit times).
Methods: month(s)|cocomo|hour(s)|commit(s).
May be multiple comma-separated values.
Alters `--loc` default to imply 'ins' (COCOMO) or
'ins,del' (hours).
-R, --recurse Recursively find repositories & submodules within <gitdir>.
-n, --no-regex Assume <f> are comma-separated exact matches
rather than regular expressions [default: False].
NB: if regex is enabled ',' is equivalent to '|'.
-s, --silent-progress Suppress `tqdm` [default: False].
--warn-binary Don't silently skip files which appear to be binary data
[default: False].
-e, --show-email Show author email instead of name [default: False].
--enum Show row numbers [default: False].
-t, --bytype Show stats per file extension [default: False].
-w, --ignore-whitespace Ignore whitespace when comparing the parent's
version and the child's to find where the lines
came from [default: False].
-M Detect intra-file line moves and copies [default: False].
-C Detect inter-file line moves and copies [default: False].
--ignore-rev=<rev> Ignore changes made by the given revision
(requires `--loc=surviving`).
--ignore-revs-file=<f> Ignore revisions listed in the given file
(requires `--loc=surviving`).
--format=<format> Table format
[default: pipe]|md|markdown|yaml|yml|json|csv|tsv|tabulate.
May require `git-fame[<format>]`, e.g. `pip install git-fame[yaml]`.
Any `tabulate.tabulate_formats` is also accepted.
--manpath=<path> Directory in which to install git-fame man pages.
--log=<lvl> FATAL|CRITICAL|ERROR|WARN(ING)|[default: INFO]|DEBUG|NOTSET.If multiple user names and/or emails correspond to the same user, aggregate git-fame statistics and maintain a git repository properly by adding a .mailmap file.
Options such as -w, -M, and -C can increase accuracy, but take longer to compute.
Note that specifying --sort=hours or --sort=months requires --cost to be specified appropriately.
Note that --cost=months (--cost=COCOMO) approximates person-months and should be used with --loc=ins.
Meanwhile, --cost=hours (--cost=commits) approximates person-hours.
Extra care should be taken when using ins and/or del for --loc since all historical files (including those no longer surviving) are counted. In such cases, --excl may need to be significantly extended. On the plus side, it is faster to compute ins and del compared to surv.
Generating CODEOWNERS:
# bash syntax function for current directory git repository
owners(){
for f in $(git ls-files); do
# filename
echo -n "$f "
# author emails if loc distribution >= 30%
git fame -esnwMC --incl "$f" | tr '/' '|' \
| awk -F '|' '(NR>6 && $6>=30) {print $2}' \
| xargs echo
done
}
# print to screen and file
owners | tee .github/CODEOWNERS
# same but with `tqdm` progress for large repos
owners \
| tqdm --total $(git ls-files | wc -l) \
--unit file --desc "Generating CODEOWNERS" \
> .github/CODEOWNERSGenerating .zenodo.json:
All source code is hosted on GitHub. Contributions are welcome.
Open Source (OSI approved):
Citation information:
- Casper da Costa-Luis (casperdcl
)
We are grateful for all 
