The plugin enables editing Jupyter notebook .ipynb files as plain text files by dynamically converting them through the jupytext command line tool.
It is a rewrite for Neovim of the jupytext.vim plugin. See History for changes relative to the original vimscript version.
The jupytext command line utility must be installed.
After following the installation instructions, the :checkhealth jupytext command can be used inside Neovim to verify the prerequisites.
Load the plugin via your favorite package manager. Lazy.nvim is recommended.
Use the following plugin specification:
{
'goerz/jupytext.nvim',
version = '0.1',
opts = {}, -- see Options
}If you are not using a package manager, copy the lua/jupytext.lua file into your runtime folder (e.g., ~/.config/nvim/). Then, in your init.lua file, call require("jupytext").setup(opts) where opts is a table that may contain any of the keys listed in Options.
The default options are:
opts = {
jupytext = 'jupytext',
format = "markdown",
update = true,
filetype = require("jupytext").get_filetype,
new_template = require("jupytext").default_new_template(),
sync_patterns = { '*.md', '*.py', '*.jl', '*.R', '*.Rmd', '*.qmd' },
autosync = true,
}The jupytext command to use. If jupytext is not on your PATH, you could set an absolute path here. Can also be set via a b:jupytext_jupytext variable.
The plain text format to use. See the description of OUTPUT_FORMAT in jupytext --help: 'markdown' or 'script', or a file extension: 'md', 'Rmd', 'jl', 'py', 'R', …, 'auto' (script extension matching the notebook language), or a combination of an extension and a format name, e.g., 'md:myst', 'md:markdown', 'md:pandoc', or 'py:percent', 'py:light', 'py:sphinx', 'py:hydrogen'.
Using format='ipynb' loads and writes the file in the original json format.
The format option may also be given as a function that calculates the format from two arguments:
path: the absolute path to the.ipynbfile being loadedmetadata: a table with metadata information from the original JSON data in the.ipynbfile. Of particular interest may be the fieldmetadata.jupytext.formatsif one wanted to implement something where paired notebooks would preferentially use the paired format instead of a common default value.
To set the format option temporarily on a per-buffer basis, set b:jupytext_format and reload the file.
Whether or not to use the --update flag to jupytext. If true (recommended), this preserves existing outputs in the edited .ipynb file. If false, every save clears all outputs from the underlying file.
This can be temporarily overridden with a b:jupytext_update variable.
The buffer filetype setting to use after loading the file, which determines syntax highlighting, etc. Can be given as a string, or more commonly as a function that returns a string after receiving three arguments:
path: the absolute path to the.ipynbfile being loadedformat: the value of theformatoptionmetadata: a table with metadata information from the original JSON data in the.ipynbfile. This should contain, e.g., the notebook language inmetadata.kernelspec.language
The default function used for this setting uses "markdown" for markdown formats, and the value of metadata.kernelspec.language otherwise. Like the previous options, b:jupytext_filetype is available to temporarily override the choice of filetype.
When editing a new (non-existing) .ipynb file, a template.ipynb file to load into the buffer. The default built-in template is for a Python notebook. While it is possible to manually edit the YAML header to change the kernel and/or language, if you primarily work with non-Python notebooks, it might be useful to set a custom template.
Patterns for plain text files that should be recognized as "syncable". If autosync=true (see below), and if, for a file matching the patterns, there also exists a file with an .ipynb extension, autocommands will be set up for jupyter --sync to be called before loading the file and after saving the file. This also periodically calls the :checktime function in the background to determine whether the file has changed on disk (by a running Jupyter server, presumably), and reloads it when appropriate.
Note that the sync_pattern only determines for which plain text files the appropriate autocommands will be set up in Neovim. The setting is independent of which Jupytext pairings are active, which is in the metadata for the .ipynb files. All linked files will automatically be kept in sync. Likewise, when editing .ipynb files directly, all linked files will be kept in sync automatically if autosync=true, irrespective of sync_pattern.
If true (recommended), enable automatic synchronization for files paired via the Jupytext plugin (the plugin for Jupyter Lab). For .ipynb files, this checks if the notebook is paired to any plain text files. If so, it will call jupytext --sync before loading the file and after saving it, to ensure all files are being kept in sync. It will also periodically check whether the .ipynb file has changed on disk and reload it if necessary, setting the autoread option for the current buffer.
When opening an .ipynb file, this plugin will inject itself into the loading process and convert the json data in the file to a plain text format by piping it through jupytext using the format set in Options.
On saving, the original .ipynb file will be updated by piping the content of the current buffer back into jupytext. With the default update setting, this will keep existing outputs and metadata in the notebook. When saving the buffer to a new file, it will be converted if the filename has an .ipynb extension. Otherwise, the buffer will be written unchanged.
While the Jupytext project provides the command line utility jupytext used by this plugin for on-the-fly conversion between .ipynb and plain text formats, its primary purpose is to provide a plugin for Jupyter to pair .ipynb files with one or more text files that are easier to manage in version control.
The intent of the original jupytext.vim plugin was to edit .ipynb files not paired in such a way, and not loaded in an active Jupyter session. With this rewritten version of the plugin, jupytext.nvim now supports editing .ipynb files with Neovim if they are paired in Jupyter, and, at least in principle, even while the notebooks are actively loaded in a running Jupyter server. For this to work, the autosync option must be set to true (default, see Options). This automatically handles the update of any paired files and watches for modifications of the file on disk while it is being edited. Saving a paired file while autosync = false will unpair it.
Even though editing files that are also actively loaded in Jupyter works, it might still be preferable to close the file in Jupyter first. The support in Jupyter for detecting external changes is not quite as good. You will have to manually reload files after saving them in Neovim. In the future, it might be possible to combine the Jupytext plugin for Jupyter with its real-time-collaboration plugin, which would alleviate this concern.
During development, make can be used locally to apply the Lua code style, generate the documentation, and run the tests. See make help for details.
Pushing commits to GitHub verifies the code formatting, the tests, and that the documentation is up-to-date via GitHub Actions.
The documentation for the plugin is maintained in this README file. This must be kept in sync with the vim help format. Running make doc locally regenerates the Vim help file from the current README to ensure this. The conversion relies on pandoc and a number of custom filters adapted from kdheepak/panvimdoc. See its documentation for details on the recommended markdown syntax to use.
GitHub Actions will check that the README and the Vim help file are in sync.
This plugin uses the plenary.nvim test framework. Tests are organized in tests/*_spec.lua files, and can be run by executing the ./run_tests.sh script. This also happens automatically on GitHub Actions with each push.
- Added: ability to create new
.ipynbfiles. These are created from a template file that can be configured via thenew_templateoption.
Initial release; Rewrite of jupytext.vim. The new plugin targets Neovim and has been rewritten in Lua to avoid restrictions of the old plugin:
- Avoid the use of temporary files in the same folder as the
.ipynbfiles potentially clashing with paired scripts. - Added support for obtaining the notebook language from its metadata, enabling use of the "script" and "auto" format.
- Added support for automatic synchronization with paired files.