Joel Grus does not like Notebooks: https://docs.google.com/presentation/d/1n2RlMdmv1p25Xy5thJUhkKGvjtV-dkAIsUXP-AL4ffI/edit?pli=1#slide=id.g362da58057_0_1

One of the best talks ever

@datajoely I'm on the vi/vim/nvim side of things, and currently exploring plugins (e.g. quarto-nvm, molten-nvim, iron.nvim, vim-slime). Further integration options for Kedro with nvim would be appreciated. 🙏 However, I can see definitely see Marimo appealing to a wider audience that are concerned about Jupyter notebooks. If you write that plugin HMU and I'll try it out and tell my colleagues about it.

I’m tempted to write a Kedro to Marimo convertor

I've talked with Akshay about this! I also went to a hack night last week to build it, but I ended up spending all my time setting up a pixi project instead 🤣

epic talk indeed! I’d love to see an updated version of it that separates what are Jupyter-specific flaws, what has been already addressed, and what is common of all notebook-like interfaces

Hi! Joined this slack because I found this thread online, I'd love to see kedro and marimo integrated more closely, what would a kedro to marimo convertor look like? @datajoely

Hi @Luis Chaves Rodriguez I actually think for 80% of pipelines it would be pretty similar Easy parts • Our

Pipeline

class should map pretty nicely to their notebook DAG • We’d have to use of

DataCatalog

to handle data management but I think that would be a good addition to something not provided out of the box in Marimo. Harder parts • I have no idea how we’d need to build the reverse, I’d like to say there’s a point where you’d want to move a Marimo prototype to a robust Kedro codebase • If you’re using Kedro lifecycle hooks to extend functionality that would be difficult

I guess, we could start smaller, like integrating kedro+marimo more nicely. I see kedro as heavily focus on mainstream DS/ML so they support pandas more than polars, ipynb more than marimo, etc One perhaps low hanging fruit that I see is easier integration of the catalog (perhaps into the DB tab of marimo), does marimo have support for plugins as far as you know?

I haven’t looked into it, but my guess it’s it’s quite easily reconstructed from their saved json format

Marimo `Cell`s have `def`s and `ref`s that correspond to outputs and inputs, respectively. For starters, I think you can construct the DAG from that.

I see kedro as heavily focus on mainstream DS/ML so they support pandas more than polars

we've had good Polars integration for a while! but it's true that the tutorial and most examples are written in pandas