Poirot helps you investigate your repositories. Give him a set of clues (e.g. strings or regular expressions) and he will report back any place they appear in your repository's revision history.
When used as a pre-commit hook, Poirot can warn you if you're about to commit something you might not intend (think passwords, private keys, tokens, and other bits of sensitive or personally identifiable information).
Poirot began as a fork of CFPB's fellow gumshoe, Clouseau.
- Running Poirot from the Command Line
- Running Poirot as a Pre-Commit Hook
- Setting up Poirot as a DC Government Employee
- Getting Involved
- Python 2.7 or 3.3+
- a UNIX-based OS (e.g. Mac or Linux) or a UNIX-y shell on Windows (e.g. Cygwin, Babun, or Git-Bash). It will not work with the default Windows Command Prompt (cmd).
Poirot uses these Python packages:
- Jinja2 to format its console output
- tqdm to display a progress bar
- regex to allow for POSIX ERE regular expressions
- requests to read remote pattern files
Poirot is available on PyPi and can be installed with pip as:
pip install poirot
You may want to install it in a virtual environment, unless you plan on using Poirot in a global commit hook.
In that case, you will have to ensure that you have done a global pip install for any Python versions you are using. E.g., if you want to run it on Python 2.7, 3.3, and 3.5 installed, install Poirot as follows:
pip2.7 install poirot pip3.3 install poirot pip3.5 install poirot
Running Poirot from the Command Line
To invoke Poirot and see his findings, call him from the command line with
poirot and the following optional arguments:
--url: The repository's URL, e.g.
firstname.lastname@example.org:DCgov/poirot.git. When included, you will be given the choice to clone or pull from the remote URL. Default value: none.
- --dir: The local path to your repository's base directory or the directory you would like to clone or pull to. Default value: the current working directory.
- --term: A single term or regular expression to search for. Default value: none.
- --patterns: The path to a .txt file with strings or regular expression patterns, each on its own line. These can be the file's URL or its relative or absolute local path. You can give a comma-separated list of pattern files, if you wish to include more than one. Default value: default.txt.
--staged: A flag, which when included, restricts search to staged revisions. This is helpful, along with
--dir, as part of a pre-commit hook.
--revlist: A range of revisions to inspect. Default value: The last commit (i.e.
--stagedis not included, otherwise none.
- --verbose: A flag to output verbose, colorful output and pattern-match highlighting. The GIF above gives an example with --verbose included.
- --before: Date restriction on revisions. Default value: none.
- --after: Date restriction on revisions. Default value: none.
- --author: Authorship restriction on revisions. Default value: none.
- --output: File to output results as JSON. Default value: none.
The most basic command Poirot will accept is:
That will search the current git directory's last commit (i.e.
HEAD^!) for the patterns in the default pattern file.
To specify one or more different patterns files (each separated by a comma), do this instead:
The --patterns option also allows files accessible over HTTP, like this one here:
To search for a single term (like
Say you want to search for
password in the whole revision history of all branches. Then do:
poirot --term="password" --revlist="all"
You can further restrict the set of revisions Poirot looks through with the
author options (which correspond to the same flags in git). E.g.:
poirot --term="password" --revlist=40dc6d1...3e4c011 --before="2015-11-28" --after="2015-10-01" --author="email@example.com"
Perhaps you don't have the repository available locally or you would like to update it from a remote URL. Just add the
url to your command and it will allow you to clone or pull to the current folder.
poirot --url https://github.com/foo/baz.git --term="password"
You can also specify a different directory than the current one with
dir. The following command will clone/pull to the folder
thisotherfolder, which sits inside of the current directory. If it does not yet exist, it will be created.
poirot --url https://github.com/foo/baz.git --term="password" --dir="thisotherfolder"
To search changes that have been staged for commit, but not yet committed, use the
poirot --term="password" --staged
Running Poirot as a Pre-Commit Hook
By setting up a pre-commit hook to run Poirot, you can have Poirot automatically run whenever you try to commit changes from the command line.
Poirot will then check these staged changes for whatever patterns you want. If there are any matches, you will have the option to cancel or go ahead with the commit. Then you can fix anything amiss and re-commit.
For a Single Repository
To set up a pre-commit hook for a particular repository, first install Poirot and then run the following from the repository's root directory:
curl https://raw.githubusercontent.com/DCgov/poirot/master/pre-commit-poirot > .git/hooks/pre-commit-poirot chmod +x .git/hooks/pre-commit-poirot echo '.git/hooks/pre-commit-poirot -f \"\" -p \"\"' >> .git/hooks/pre-commit chmod +x .git/hooks/pre-commit
-p in the second to last line are flags for patterns folder and a comma-separated list of pattern files, respectively. These let you use patterns other than the default, if you would like, by providing their absolute path or URL.
For example, you could change the flag values to
-f "/Users/myusername/Documents/poirot-patterns" -p "https://github.com/DCgov/poirot-patterns/blob/master/financial.txt, https://github.com/DCgov/poirot-patterns/blob/master/id.txt".
You can either edit that line before you run it, or edit it after with:
My advice is to fork the poirot-patterns repository and download it to your computer. Then add the absolute path to that folder as the
If you go ahead with setting up a patterns folder, then you can easily add, delete, or modify the pattern files without having to keep re-editing the commit hook.
As an aside, if you ever want to commit without running the hook, just use:
git commit --no-verify
For All Repositories
To set a Poirot pre-commit hook for all your new repositories, you can add it to your default template with the init.templatedir configuration variable. Then, whenever you
git init a repository, Poirot will be set to run. The following code will do that for you:
mkdir -p ~/.git_template/hooks git config --global init.templatedir '~/.git_template' curl https://raw.githubusercontent.com/DCgov/poirot/master/pre-commit-poirot > ~/.git_template/hooks/pre-commit-poirot chmod +x ~/.git_template/hooks/pre-commit-poirot echo '.git/hooks/pre-commit-poirot -f \"\" -p \"\"' >> ~/.git_template/hooks/pre-commit chmod +x ~/.git_template/hooks/pre-commit
As in the Single Repositories case, the
-p flags in the second to last line will let you use pattern files other than the default. If you don't care about that, you can run it as is.
For existing repositories, you can re-run
git init in the repo. Running
git init will not overwrite things that are already there. It will only add new template files (e.g. this hook).
As in the above section on Single Repositories, I recommend that you start out by setting up a patterns folder on your computer. You can fork and download the poirot-patterns repository to get started.
Using a patterns folder avoids most instances that would make you want to make subsequent edits to your global pre-commit hook.
If you do decide to make a change after you have already applied the global pre-commit hook to a repository, you will need to delete the repository's existing pre-commit hook and re-run
git init. To do that, run the following from the root of the repository:
rm ~/.git/hooks/pre-commit git init
Setting up Poirot as a DC Government Employee
Poirot is part of the standard DC Government open source development toolkit. Once you have Python and pip set up on your computer, run the following from the command line:
pip install poirot civicjson groupthink
This installs Poirot, in addition to the Civic.JSON-CLI (a Python package that helps you document your code projects with a civic.json file), and Groupthink (another Python package that will allow you to manage project set-up scripts).
Next, install the dcgov-cli scripts by running:
groupthink install dcgov
When you want to start an open source DC Government project, go to your new project's directory and run:
This will download the standard DC Government open source license and contributing files, set up Poirot as a pre-commit hook to run whenever you attempt to commit changes to the project from the command line, and give you the option to install a civic.json file.
Hey! Glad you're interested in getting involved, whether by flagging bugs, submitting feature requests, or otherwise improving Poirot.
To get you oriented, there are three project repositories to be aware of:
- This one here, which contains the Poirot Python package.
- DCgov/poirot-patterns, where we're compiling boilerplate pattern files.
- DCgov/poirot-test-repo, which we're running the tests on. If you check out the test directory in this repository, you will find that DCgov/poirot-test-repo has been added as a submodule.
If you're going to modify a Poirot fork and submit pull requests, be sure to add tests to validate your changes.
Running Unit Tests
Once you've forked or cloned Poirot, you can run the unit tests with:
python setup.py test
To test multiple Python versions at once (current we aim to support 2.7, 3.3, 3.4, and 3.5), you will need each installed in your environment (I recommend using pyenv).
The tox package will let you run the tests in one go. Install tox with pip or easy_install, then simply run it with:
It uses the tox.ini file to know what to do.