It also requires no additional syntax for reactive features, which is nice for non-devs.
The most common issue with these kind of reactive environments, especially python ones, is the behaviour around long-running calculations.
Last thing, the 'no hidden state' claim can fall down in pratice esp with python's inherent mutability, and e.g. file systems/IO when they're involved.
As you can probably tell, I'm a fan of marimo. I do most of my work in it. But I'm always interested in new ideas around the concept, and this seems like something you put a lot of effort into so wanted to share some feedback.
In signals reactivity the same important rule applies that formulas/derivations shouldn’t mutate state of objects and rather should create a new or copy of the object (like lists, dicts or np.arrays). I think this is the same marimo issue you are talking about. :)
Hey guys! After I tried to reintroduce Signals Reactivity state management (inspired by Angular and SolidJS) to the Python community with my reaktiv package, I like to go one step further to reimagine how Python Notebooks (Jupyter Notebook, Marimo, …) could look like, if Signals gets used as the core mechanism for reactive data pipelines.
The clue here: each cell in the spreadsheet represents a signals primitive under the hood! Say hello again to Signal, Computed, Effect, Linked and Resource!
So effectively you are using the spreadsheet interface as a canvas to model your business logic around these core primitives. Each cell represents a data pipeline execution.
I am working in the signal processing field and like seeing dense data and see how parameters changes and affects the outcome. I also always liked how spreadsheets used data as part of their UI instead of writing an UI on top of the data.
Hopefully some data folks appreciate what I have done here. :)
Combining Frontend Tech with Data Stuff in Python.
I hope this takes off. A decade ago, there was something called Dirigible - "Excel, but with python". I've been hoping for a replacement since that product was retired.
Thanks! :) Exploring in spreadsheets is nice because, like with Signals, you are starting with small basic calculations / cells and can derive them further without affecting the previous calcs. I think there is a term for this: orthogonal programming?
Nice! Love the dependency graph! I have a story how an excel spreadsheet ended up mispricing an asset by $50M, due to the formulas being hard to review. Crazy how much can go wrong in excel.
I'll give this a spin and check out the reaktiv framework too!
Also really fun to see others try different frontend approaches for data work. I have a SQL canvas: https://github.com/aleda145/kavla
Currently no reactivity though, so I will absolutely take a peek at how you've implemented signals.
Dependency views: https://docs.marimo.io/guides/editor_features/dataflow/
It also requires no additional syntax for reactive features, which is nice for non-devs.
The most common issue with these kind of reactive environments, especially python ones, is the behaviour around long-running calculations.
Last thing, the 'no hidden state' claim can fall down in pratice esp with python's inherent mutability, and e.g. file systems/IO when they're involved.
As you can probably tell, I'm a fan of marimo. I do most of my work in it. But I'm always interested in new ideas around the concept, and this seems like something you put a lot of effort into so wanted to share some feedback.
In signals reactivity the same important rule applies that formulas/derivations shouldn’t mutate state of objects and rather should create a new or copy of the object (like lists, dicts or np.arrays). I think this is the same marimo issue you are talking about. :)
I need to add this info to the tutorial.
The clue here: each cell in the spreadsheet represents a signals primitive under the hood! Say hello again to Signal, Computed, Effect, Linked and Resource!
So effectively you are using the spreadsheet interface as a canvas to model your business logic around these core primitives. Each cell represents a data pipeline execution.
I am working in the signal processing field and like seeing dense data and see how parameters changes and affects the outcome. I also always liked how spreadsheets used data as part of their UI instead of writing an UI on top of the data.
Hopefully some data folks appreciate what I have done here. :)
Combining Frontend Tech with Data Stuff in Python.
Also congrats to SolidJS releasing v2!
reaktiv-workbook: https://github.com/buiapp/reaktiv-workbook
reaktiv: https://github.com/buiapp/reaktiv
Intuitively I like the idea of exploring the space between notebooks and spreadsheets.
I'll give this a spin and check out the reaktiv framework too!
Also really fun to see others try different frontend approaches for data work. I have a SQL canvas: https://github.com/aleda145/kavla
Currently no reactivity though, so I will absolutely take a peek at how you've implemented signals.
Can I export to PDF please?