Feature Request: Create New Folder w/ venv

[jthomasmock]

Positron Version:

Steps to reproduce the issue:

As someone who largely uses virtual environments, I frequently hit the "Python support requires module ipykernel. Install?" each time I open up a new project/folder.

The VS Code Python extension has a built in way to create environments via Venv or Conda: https://code.visualstudio.com/docs/python/environments#_creating-environments

This helps with hooking up all the plumbing so that VSCode/Positron gets users into a good starting state of the virtual environment is created, recognized, and respected.

In Positron, it might be nice to build on that concept and support creating a folder with some predefined R/Py structure:

• Pre-select a version of R/Python
• Option to use Venv/Conda with Python or renv with R, and maybe indicate that IPykernel will be pre-installed into virtual environment?
• Don't have to use a .RProj or some other project dotfile, but initialize the virtual environment, preferably pre-install IPykernel (and tell the user this has happened) and preferably create a minimal .gitignore that ignores the virtual environment folder.

What did you expect to happen?

Were there any error messages in the output or Developer Tools console?

[jthomasmock]

Also related to: #1371

[jthomasmock]

There's also some interesting discussion from the python-extension folks at: microsoft/vscode-python#19314

[juliasilge]

In our Python OSS sync meeting this week, the following list of jobs/tasks was put together about what folks want to do when they are creating a new workspace:

• New project
• With a virtual environment named .venv
• That already contains IPykernel
• With a gitignore for .venv
• With the selected interpreter set as the .venv

This should happen before even opening the folder, so the experience feels very seamless.

[jthomasmock]

Wow, looks great! Minor nit, but probably default to check for Install ipykernel, since it's essentially required for a fresh environment?

Possibly out of scope...
but do you want to include an option for a "Project type" in the future or is that out of scope for now?

[seeM]

Agreed on the ipykernel checkbox default. It's a bit of a weird one because you can't actually use the environment in Positron without ipykernel installed so maybe a checkbox isn't the right component. At the same time, we don't want to install something into a user's environment without their permission. Maybe a checkbox with a note?

I've personally been focused on how we integrate Python environment management into New Folder. I think "project type" is important, but not sure if we tackle that at the same time or later – perhaps @petetronic can comment?

[petetronic]

I think project type will be an earlier question.

Thanks for pushing this along @seeM, maybe we could rearrange the order of questions such that you pick an interpreter first, and then if that isn't already a venv etc., then offer the section to create one as a secondary consideration?

[juliasilge]

We presented this UI mockup in the monthly Python OSS sync and folks were so excited about having good tooling around this! Here is some feedback:

• Folks strongly want the ipykernel installation to be default "yes" so you would need to opt out of installing ipykernel if you really want that for some reason.
• It is possible and in fact a good practice to start from an environment with a venv and make a new venv for a new project, i.e. the original environment is used as a template and copied but the original venv is not used. We'd like either starting from a base interpreter or a "previously configured" interpreter to then lead to making a new venv.
• Being possible to skip making a new venv is acceptable but folks strongly want making a new venv for a new project opt-out. This is an area where being more opinionated could really help our users. We particularly talked about Posit customers on Workbench and how support in this area would be helpful.
• Folks want the ability to have a template-like experience where they have, say, a default data analysis or ML requirements.txt and then can make new but separate projects from that template.
• Pretty much everything in this screenshot can be confusing to new users. Can we add some nice tooltips / hotspots that users can click on to get more info, like a brief description of what a virtualenv is and is used for, same for interpreter, what is "location" in this context, etc.

[isabelizimm]

Few thoughts:

• If project type will be earlier, will it still be necessary to select a language (eg, if I choose "Python package")? I wonder if that could be eliminated 🤔
• Would "Location" be automatically filled? Also, is this the location of a .venv file, or the path to the interpreter?

Overall, I like the look of this for a simple data science Python folder!

[jthomasmock]

cc @sharon-wang

[jthomasmock]

JupyterLab now has a "new environment" wizard, but is not necessarily tied to a new folder.

https://blog.jupyter.org/python-environment-management-in-jupyterlab-desktop-3b119c5811d9

[sharon-wang]

This issue has been divided into:

• Project Wizard: Python project environment step - Venv #2838
• Project Wizard: Python project environment step - Conda #2839
• Project Wizard: Python project environment step - existing interpreter #2840
• Project Wizard: create renv::init command #3127