Strongly opinionated python project management.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
bin
docs
medikit
tests
.coveragerc
.gitignore
.style.yapf
.travis.yml
MANIFEST.in
Makefile
Projectfile
README.rst
RELEASE.rst
TODO.rst
classifiers.txt
readthedocs-conda.yml
readthedocs.yml
requirements-dev.txt
requirements.txt
setup.cfg
setup.py

README.rst

✚ medikit ✚

Strongly opinionated python 3.5+ project management.

Continuous Integration Status Coverage Status Documentation Status License Status

Medikit is the first-aid toolkit to manage your project's boilerplate, like package files, versions, config, test suite, runners, ...

This package helps you create python (or not) source trees using best practices (or at least the practices we consider as best for us) in a breeze.

Don't worry about setting up git, a makefile, usual project targets, unit tests framework, pip, wheels, virtualenv, code coverage, namespace packages, setup.py files ...

Medikit's got you covered on all this, using one simple and fast command.

Install

Before installing the package, you must make sure that pip and virtualenv are installed and available to be used in your current environment.

pip install medikit

Now, you may want to bootstrap a python package source tree.

mkdir my.awesome.pkg
cd my.awesome.pkg
medikit init .

You're done with the bootstrap. You can now run:

make install
make test
git commit -m 'Damn that was fast ...'

Happy?

Update

If you change the Projectfile content, or update the library, you will need to run the generator again.

medikit update

To better control what changes are made, I suggest that you run it on a clean git repository, then look at the dofferences using:

git diff --cached

You can then commit the generated changes.

Gotchas

As the headline says, we have made strong opinionated choices about how a project tree should be organized.

For example, we choose to use make to provide the main project entrypoints (install, test). We also choose to use git. And pytest. And to put root package in the project root (as opposed to a src dir or something like this). Etc.

For beginners, that's a good thing, because they won't have to ask themselves questions like "What should I put in setup.py ?" or "Should I create a «src» dir or not ?". For more advanced users, it can be either a good thing if you agree with our choices, or a bad one ...

Architecture

Medikit uses a single configuration file in your project, called Projectfile.

This file contains all the specific of your project:

  • What features you wanna use.
  • The customizations of those features
  • The additional event listeners (more on this later) you need.
  • The eventual pipelines that you need.

At its heart, medikit uses an "event dispatcher" model.

An update will dispatch a "medikit.on_start" event, and features that you required can listen (react) to this event by adding jobs to run in response. They also can dispatch their own events.

As a result, you'll get your projects files updated, that will be a combination of all the events listeners executed.

It means two things:

  • Unlike usual project templates and generators, it can both bootstrap and update your project, as best practice evolves.
  • It's not a dependency of your project. Once it has run, you can forget it. Either you choose to maintain your project assets with it and you'll need it installed while updating, or you can remove it and just keep the generated files.

F.A.Q

  • I'm using PasteScript, isn't that enough?
    • PasteScript with the basic_package template will only generate a very minimalistic tree, while we install a few tools and generate more boilerplate than it does. The fact is, we were using it before but still had a lot of repeated actions to do then, and the exact aim of this project is to automate the whole. Also, PasteScript cannot update a project once generated, while we do.
  • Should I use it?
    • You're a grown man, right?
  • Is it stable / production ready?
    • Not really relevant to this project, as it's more a development tool than something you'll use in operations. However, please note that on some points and until version 1.0, we will tune things and change the way it works to find the most flexible way to operate. Thus, if you relly on a specific implementation, updates may break things. The good news is that you'll be able to review changes using git diff --cached, and either rollback or report issues saying how much you're disappointed (and why, don't forget the why, please).
  • Can I contribute?
    • Yes, but the right vs wrong choices decision is up to us. Probably a good idea to discuss about it (in an issue for example) first.
  • Can you include feature «foo»?
    • Probably, or maybe not. Come on github issues to discuss it, if we agree on the fact this feature is good for a lot of usages, your patch will be welcome. Also, we're working on a simple way to write "feature plugins", so even if we don't agree on something, you'll be able to code and even distribute addons that make things work the way you like.
  • Do you support python 3?
    • Of course, and for quite some times we decided to only support python 3, as we think the "10 years incubation period" we just had is a sufficient maturation period to just forget about python 2.

License

License Status