Skip to content

Latest commit

 

History

History
121 lines (86 loc) · 4.17 KB

README.md

File metadata and controls

121 lines (86 loc) · 4.17 KB
Raw

Fakegir: Bring autocompletion to your PyGObject code!

fakegir is a tool to build a fake python package of PyGObject modules.

After inserting this package in your editor's python path, you will get autocompletion for every PyGObject module!

By default the package is saved in ~/.cache/fakegir/. This is the directory you want to add to your Python PATH. Of course, you shouldn't put this package in your global PYTHONPATH or your PyGObject applications will stop working immediatly ;)

Requirements

You'll need python-lxml to parse gir files.

In order to get every required gir file, install the libgirepository1.0-dev package

If you want autocompletion for additional packages, you can find it with apt-file :

$ apt-file search ".gir" | grep -i unity
libunity-dev: /usr/share/gir-1.0/Unity-6.0.gir
libunity-webapps-dev: /usr/share/gir-1.0/UnityWebapps-0.2.gir
libunity-webapps-dev: /usr/share/gir-1.0/UnityWebappsRepository-0.2.girV
$ sudo apt-get install libunity-dev

Usage

  • ./fakegir.py
  • Add ~/.cache/fakegir/ to the PYTHONPATH of your favorite editor
  • ????
  • Profit !

Don't forget to run fakegir when you install new gir files or upgrade your distro.

Depending on your editor, it is highly recommended that you build fakegir modules without the docstrings which can make autocompletion much slower or even freeze your computer, with you still with to build fakegir with docstrings you can run it with the WITHDOCS variable::

WITHDOCS=1 ./fakegir.py

By default fakegir scans /usr/share/gir-1.0 for gir files. You can override this by setting a GIRPATH environment variable. You can set more than one directory, separated by colons.

Gedit support

You can use Gedit Fakegir Loader by @gabrieltigre: https://github.com/gabrieltigre/gedit-fakegir-loader

VS Code Support

For Pylance language server (which is the default), you can add extra paths with this setting:

"python.analysis.extraPaths": [
     "/home/USERNAME/.cache/fakegir/"
]

For Jedi language server, you can add extra paths with this setting:

"python.autoComplete.extraPaths": [
     "/home/USERNAME/.cache/fakegir/"
]

Vim support

Vim support is a bit tricky for the moment. There are several options but these may or may not work with your Vim setup.

For a basic vim setup, you can insert the snippet below to your vimrc:

if has('python')
py << EOF
import os.path
import sys
import vim
sys.path.insert(0, os.path.join(os.path.expanduser('~'), '.cache/fakegir/'))
EOF

A lot of Vim users use YouCompleteMe for autocompletion which itself uses Jedi for Python completion. Getting Fakegir to work with a bit trickier. The first option is to run the provided script build-jedi-cache.sh. This script is from a fork of Fakegir on github I merged back into my branch. I haven't witnessed this script actually working with the current version of Jedi but I'll leave it here in case someone manages to get something out of it. (Please make a Pull Request if you do).

Also it's worth mentioning that:

  • build-jedi-cache.sh requires docopt module to run. On Debian/Ubuntu it is located in python-docopt and python3-docopt packages for Python 2 and Python 3, respectively.

  • Python interpreter in build-jedi-cache.sh defaults to python, which usually points to Python 2. Use -3 CLI option to make the script use python3 instead, or -p /path/to/python to force an absolute interpreter path.

The other option which I use is to create a virtualenv and copy the fakegir gi package into it:

mkvirtualenv fakegir
cdvirtualenv
lib/python*/site-packages
cp -a ~/.cache/fakegir/gi .

You'll have to activate this virtualenv before editing your GObject program with vim which makes it a bit more cumbersome but this method is the only one I could get autocompletion with vim and Jedi. Note that creating a Python3 virtualenv doesn't seem to work and that the first autocompletion can be quite slow while Jedi is building the cache. (Fakegir modules, while containing no code at all are very large).