The modification history of a program is represented by its version control log and can be a huge source of interesting insights, in particular if combined with static source code analysis. This program analyses a git history and provides multiple modules (currently: one) to compute fascinating insights.
This is a standard gradle project, i.e. you build it via
gradle build
Note that in addition to the usual output in the build/ directory a standalone executable file git-analyzer will be
created in the code's root directory as well (and removed on gradle clean).
Start the analyzer without arguments to see the available options:
$ # gradle build
$ ./git-analyzer
Value for option --directory should be always provided in command line.
Usage: git-analyzer options_list
Options:
--directory, -d -> Source code directory (always required) { String }
--period, -p [128y] -> Period, e.g. 1y, 2w, 2w3d, ... { String }
--analysis, -a [PackageExperts] -> Analysis to execute { Value should be one of [packageexperts] }
--help, -h -> Usage info
By default, you only need to provide a directory pointing to a git repository. For example, to gather statistics for JUnit5 using the current default analyzer, write
$ git clone https://github.com/junit-team/junit5.git /tmp/junit5
$ ./git-analyzer --directory /tmp/junit5
{
<a lot of json output>
"platform.tooling.support.tests" : {
"Sam Brannen <[email protected]>" : 1,
"Juliette de Rancourt <[email protected]>" : 2,
"Christian Stein <[email protected]>" : 76,
"JUnit Team <[email protected]>" : 85,
"Marc Philipp <[email protected]>" : 113
},
"standalone" : {
"Christian Stein <[email protected]>" : 1,
"Sam Brannen <[email protected]>" : 3,
"JUnit Team <[email protected]>" : 9
}
}
The package platform.tooling.support.tests has been edited by five different accounts and Marc Philipp had the most
commits (113) and could be considered the technial expert; note that this analysis might be skewed by team accounts such as the "JUnit Team", so be aware of that.
By specifying --period you can restrict analysis to a specific period, e.g. last 100 days via --period 100d, last 3
months via --period 3m etc.
This repository also provides an exemplary script which shows how to use the JSON output in combination which modern
UNIX tools like jq and fzf. See scripts/query.sh:
Splits up the code base by packages and determines experts via their respective number of commits. Currently, does not take renamed files into account.
- Write tests
- Build via GitHub actions
- Add more modules, e.g. files often changed together, most changed files per package, ...
The source code is licensed under the Apache license