Support organizing imports in jupyter cells #51
Comments
|
Cool, I'm sure we can support this but I have to look into how it's done. |
|
The team over at microsoft/vscode-black-formatter have their whole-notebook support blocked by pygls still being on LSP v3.16. Looks like @charliermarsh you are too. The closest related issue over at |
|
@guyrosin organize imports doesnt work for me, how did you get it to work for you? |
|
If you install the isort vscode extension you can use its Organize Imports action |
|
exactly what I did, but its the isort extension doing the work. when I choose ruff it doesnt do anything |
|
Yep, that's what this issue is all about :) |
|
I think this information (new VSCode release) could be relevant to this issue:
|
|
We'd love to see this too - we're keeping isort around just for this atm. |
|
We very much want to add this, but we're waiting on support from the library we use for our LSP server (pygls). See: openlawlibrary/pygls#311. |
|
I see this has been marked as closed but how do we sort imports on save with |
|
This used to work, but I think something regressed in VS Code. We filed an issue: microsoft/vscode#216653. |
## Summary This PR adds support for Jupyter Notebook. It requires client support for LSP 3.17 which contains the [Notebook support](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#notebookDocument_synchronization). ### Implementation #### Context * `Document`: LSP type representing a text file (Python file for Ruff). * `TextDocument`: `pygls` representation of the LSP `Document`. This is an abstraction created from a `Document` which provides some useful methods like getting the file path, source code, etc. * New in 3.17: `NotebookDocument` type was added to represent a Notebook which consists of a list of cells (`NotebookCell`). Note that these are all LSP types coming from `lsprotocol`. * In `pygls`, a Notebook cell is represented as a text document (`TextDocument`). There are methods provided by `pygls` to get the object: * `get_text_document` - Returns a `TextDocument` which either represents a Python file or a Notebook cell * `get_notebook_document` - Returns a `NotebookDocument` either using the Notebook URI or a cell URI. For cell URI, it returns the `NotebookDocument` containing the cell. #### Document A new `Document` type was created to facilitate the implementation. This represents either a Python file, a Notebook or a Notebook cell. There are various constructor methods which should be used to create this type: * For a URI representing a Python file, use either `from_uri` or `from_text_document`. * For a URI representing a Notebook file, use either `from_uri` or `from_notebook_document`. * For a URI representing a Notebook cell, use either `from_cell_or_text_uri` or `from_notebook_cell`. #### Notebook JSON Ruff expects the source content of a Notebook file to be in JSON format following the [Notebook format specification] but the protocol uses it's own abstraction and doesn't store the JSON format. This means that we need to create a JSON string representing the Notebook from the available information. This doesn't need all the information as Ruff only uses the cell source and version information. So, we create a minimal JSON string representing the Notebook document and pass it to Ruff. <details><summary>An example JSON string representing a Notebook Document:</summary> <p> ```json { "metadata": {}, "nbformat": 4, "nbformat_minor": 5, "cells": [ { "cell_type": "code", "metadata": null, "outputs": [], "source": "import random\nimport math" }, { "cell_type": "code", "metadata": null, "outputs": [], "source": "try:\n print()\nexcept ValueError:\n pass" }, { "cell_type": "code", "metadata": null, "outputs": [], "source": "import random\nimport pprint\n\nrandom.randint(10, 20)" }, { "cell_type": "code", "metadata": null, "outputs": [], "source": "foo = 1\nif foo is 1:\n msg = f\"Invalid foo: {foo}\"\n raise ValueError(msg)" } ] } ``` </p> </details> **We need to pass in every cell including the markdown cell to get an accurate information like the cell number.** For the cell document kind, the source value is a JSON string containing just a single code cell. This is required as code actions and formatting work at both cell and notebook level. ### Configuration For VSCode users, the `notebook.*` configuration is used to run the formatter or code actions on save: ```jsonc { // Enable formatting the entire Notebook on save "notebook.formatOnSave.enabled": true, // Run the enabled code actions on the entire Notebook on save "notebook.codeActionsOnSave": { "source.fixAll": true, "source.organizeImports.ruff": true }, } ``` The way the above settings work in VSCode is that the editor runs the actions in parallel for every cell. This has the illusion that it was run on the entire Notebook. The commands defined by us (`Ruff: Organize imports` and `Ruff: Fix all auto-fixable problems`) are run on the entire Notebook at once. This is important because in the latter case the `ruff` command is invoked `n` number of times where `n` is the number of cells while for the former it's run only once. ### Commands #### Builtin * `Ruff: Organize Imports`: Works at Notebook level * `Ruff: Fix all auto-fixable problems`: Works at Notebook level #### VSCode specifc * `Format Cell`: Formats the current cell * `Notebook: Format Notebook`: Formats the entire Notebook by running the formatter for every cell * `Organize Imports`: Runs the `source.organizeImports` code action on every cell in parallel * `Fix All`: Runs the `source.fixAll` code action on every cell in parallel ## Feature checklist - [x] Code actions - [x] Organize imports - [x] Fix all - [x] Each fixable diagnostics - [x] Disable rule comment - [x] Code action resolve - [x] Commands - [x] `ruff.applyAutofix` - [x] `ruff.applyOrganizeImports` - [x] `ruff.applyFormat` - [x] Diagnostics - [x] On open - [x] On close - [x] On save - [x] On change - [x] Formatting - [x] Hover ## Test Plan Manually testing for all the features mentioned above. ### How to run this locally? 1. Clone https://github.com/astral-sh/ruff-lsp and https://github.com/astral-sh/ruff-vscode in the same directory 2. Checkout this branch `git checkout dhruv/notebook` in the `ruff-lsp` repository 3. Install the requirements for both repositories 4. For `ruff-vscode`, uninstall `ruff-lsp` (`pip uninstall --yes ruff-lsp`) as we'd want to use the local version. To install the local `ruff-lsp` version in `ruff-vscode`, follow [Modifying the LSP](https://github.com/astral-sh/ruff-vscode#modifying-the-lsp). 5. Open VSCode from `ruff-vscode` directory -> "Run and Debug" section from the sidebar -> "Debug Extension and Python" config. This will then open a VSCode development session which can be used to test out the notebook features. **Test notebooks:** * Formatting: https://gist.github.com/dhruvmanila/7803e5a3b98c414505384db415a635a0 * Diagnostics, Code actions, Commands: https://gist.github.com/dhruvmanila/54c65870f167a56558d4701f57f53042 **Requires: astral-sh/ruff#7664 which was released in `v0.1.0`** fixes: #267 closes: astral-sh/ruff-vscode#256 closes: astral-sh/ruff-vscode#314 closes: astral-sh/ruff-vscode#51
The "Organize Imports" action of the isort vscode extension works for jupyter cells while ruff's Organize Imports does nothing when attempting to run in a cell.
The text was updated successfully, but these errors were encountered: