Pynvim: Python client to Neovim
Pynvim implements support for python plugins in Nvim. It also works as a library for
connecting to and scripting Nvim processes through its msgpack-rpc API.
Install
Supports python 3.10 or later.
- Installation option #1: install using uv (recommended):
- Installation option #2: install using pipx:
- Install pipx (https://pipx.pypa.io/stable/).
- Install pynvim:
pipx install pynvim
- Anytime you upgrade Neovim, make sure to upgrade pynvim as well by
running:
pipx upgrade pynvim
- Other installation options:
- See pynvim installation
documentation
for additional installation options and information.
Python Plugin API
Pynvim supports python remote plugins (via the language-agnostic Nvim rplugin
interface), as well as Vim plugins (via the :python3 interface). Thus when
pynvim is installed Neovim will report support for the +python3 Vim feature.
The rplugin interface allows plugins to handle vimL function calls as well as
defining commands and autocommands, and such plugins can operate asynchronously
without blocking nvim. For details on the new rplugin interface,
see the Remote Plugin documentation.
Pynvim defines some extensions over the vim python API:
- Builtin and plugin vimL functions are available as
nvim.funcs
- API functions are available as
vim.api and for objects such as buffer.api
- Lua functions can be defined using
vim.exec_lua and called with vim.lua
- Support for thread-safety and async requests.
See the Python Plugin API documentation for usage of this new functionality.
Known issues
As Nvim does not support
bindeval, accessing a Vimscript
dictionary from Python (via vim.vars, vim.vvars, Buffer.vars, etc)
returns a copy, not a reference. As a result, setting its fields directly will
not write them back into Nvim. Instead, the whole dictionary must be written as
one. This can be achieved by creating a short-lived temporary.
Example:
vim.vars['my_dict']['field1'] = 'value' # Does not work
my_dict = vim.vars['my_dict'] #
my_dict['field1'] = 'value' # Instead do
vim.vars['my_dict'] = my_dict #
Development
Use (and activate) a local virtualenv, for example:
python3 -m virtualenv venv
source venv/bin/activate
If you change the code, you must reinstall for the changes to take effect:
pip install .
Use pytest to run the tests. Invoking with python -m prepends the current
directory to sys.path (otherwise pytest might find other versions!):
python -m pytest
Or use uv to run tests and linting without managing a virtualenv:
uvx --with pynvim --with pytest --with pytest-timeout pytest
uvx --with pynvim --with flake8-import-order --with flake8-docstrings --with pep8-naming flake8 pynvim test
uvx --with pynvim --with msgpack-types mypy --show-error-codes pynvim test
For details about testing and troubleshooting, see the
development
documentation.
Usage from the Python REPL
A number of different transports are supported, but the simplest way to get
started is with the python REPL. First, start Nvim with a known address:
$ nvim --listen /tmp/nvim.sock
Or alternatively, note the v:servername address of a running Nvim instance.
In another terminal, connect a python REPL to Nvim (note that the API is similar
to the one exposed by the python-vim
bridge):
>>> import pynvim
# Create a session attached to Nvim's address (`v:servername`).
>>> nvim = pynvim.attach('socket', path='/tmp/nvim.sock')
# Now do some work.
>>> buffer = nvim.current.buffer # Get the current buffer
>>> buffer[0] = 'replace first line'
>>> buffer[:] = ['replace whole buffer']
>>> nvim.command('vsplit')
>>> nvim.windows[1].width = 10
>>> nvim.vars['global_var'] = [1, 2, 3]
>>> nvim.eval('g:global_var')
[1, 2, 3]
You can embed Neovim into your python application instead of connecting to
a running Neovim instance.
>>> import pynvim
>>> nvim = pynvim.attach('child', argv=["/usr/bin/env", "nvim", "--embed", "--headless"])
- The
--headless argument tells nvim not to wait for a UI to connect.
- Alternatively, use
--embed without --headless if your client is a UI
and you want nvim to wait for your client to nvim_ui_attach before
continuing startup.
See the tests for more examples.
Release
- Create a release commit with title
Pynvim x.y.z
- list significant changes in the commit message
- bump the version in
pynvim/_version.py
- Push to
master.
git push
- Make a release on GitHub with the same commit/version tag and copy the message.
- Run
scripts/disable_log_statements.sh
- Run
pipx run build
- (Validation) Diff the release tarball
dist/pynvim-x.y.z.tar.gz against the previous one.
- Fetch the previous tar.gz from https://pypi.org/manage/project/pynvim/releases/
- Unzip both.
- Unzip both.
- Diff them with
:DiffTool old/ new/ (plugin: https://github.com/deathbeam/difftool.nvim)
- Run
pipx run twine upload -r pypi dist/*
- Assumes you have a pypi account with permissions.
- Run
scripts/enable_log_statements.sh or git reset --hard to restore the working dir.
- Bump up to the next development version in
pynvim/_version.py, with prerelease suffix dev0.
Releasing with bump-my-version
bump-my-version automates the process of updating version strings, creating git commits, and tagging releases.
-
Install bump-my-version:
If you haven't already, install the development dependencies:
pip install .[dev]
-
Bump the version:
To increment the version, use one of the following commands:
- Patch release:
bump-my-version bump patch (e.g., 0.6.1 -> 0.6.2)
- Minor release:
bump-my-version bump minor (e.g., 0.6.1 -> 0.7.0)
- Major release:
bump-my-version bump major (e.g., 0.6.1 -> 1.0.0)
This command will:
- Update the
version in pyproject.toml.
- Update the
VERSION in pynvim/_version.py.
- Create a git commit with a message like "Bump version: 0.6.1 → 0.6.2".
- Create a git tag (e.g.,
v0.6.2).
-
Push changes and tags:
After bumping the version, push the commit and the new tag to your remote repository:
git push --follow-tags
License
Apache License 2.0
