jupyter-lablanguage-server-protocolpython-language-serverpython-lsp-serverjupyterlab-lsp

Why is jedi installed?

My basic understanding is that jedi was an old method to do autocompletion / hints.

The new method is LSP, and we can use an LSP server with jupyterlab through jupyterlab-lsp.

One such LSP server for Python is python-lsp-server.

There is also a jedi-language-server that, as my understanding goes, is a layer that exposes a LSP interface to jedi (the old method).

Now I want to use python-lsp-server.

I use pyenv to manage my virtual environments.

I do:

# This creates a Python 3.12 virtual environment
$ pyenv virtualenv 3.12.0 nojedi

# Basically this activates the environment
$ pyenv local nojedi

# Install packages
$ python -m jupyterlab jupyterlab-lsp "python-lsp-server[all]"

Then:

$ python -m pip list | grep jedi
jedi                      0.19.1

Why is jedi installed?

Solution

  • python-lsp-server depends on Jedi.

    From its README:

    The base language server requires Jedi to provide Completions, Definitions, Hover, References, Signature Help, and Symbols:

    The only “obsolete” thing about Jedi when it comes to text editors is having its own API. It didn’t stop being good at Python static analysis.