Testing out interest in integrating my dataframe display library built around ag-grid

[paddymul]

Hi,
I built Buckaroo to become the default method of DataFrames in Notebooks. It is built around ag-grid. I'm looking to promote and distribute my project and see if we could collaborate.

Buckaroo has two parts, the widget which works in notebooks, and a framework for building table display apps. I think you'd be interested in the table display framework. The interactive parts of buckaroo (everything attached to the status bar above the main table) probably aren't a fit for nicegui.

Check out this interactive styling gallery which shows how buckaroo can be configured with different display options. That's a WASM notebook so it takes a little bit to load.

Buckaroo makes it easy to configure ag-grid tables with the following features:

1. Completely declarative based config of ag-grid. Buckaroo doesn't pass js functions to the frontend.
2. Sparkline cell displayers including BarCharts, Area Charts, and Line Charts. These are implemented with recharts and don't require ag-grid enterprise
3. Pinned Rows that can use different displayers than the main table body
4. Cell coloring based on value
5. High performance infinite scrolling based on parquet. Buckaroo only loads 100 rows at a time, and the rest can be sent piecemeal as the user scrolls. This is sent over the wire with parquet which is much faster to serialize and deserialize than JSON (it has better types too).
6. You can play with buckaroo interactively in a notebook trying a variety of configurations, and take this config to nicegui.

Why Buckaroo might not be a fit for nicegui users:

1. Cells aren't editable. I have not pursued this feature because editting locations in a dataframe doesn't fit with the normal workflows in Jupyter notebooks.
2. Buckaroo is very capable, but adding JS functionality requires a new build of buckaroo-js-core. You can't just send in JS.
3. Styling is hardcoded to dark-mode. This is definitely possible to update.
4. Styling is generally hardcoded. Buckaroo uses ag-grid themes, adding a theme parameter would be pretty straightforward.

What would need to be built to integrate into nicegui:

1. I would probably need to make Buckaroo-js-core release as a separate npm module. Devops, but easily doable
2. Create a nicegui-buckaroo class similar

   nicegui/nicegui/elements/table.py

   Line 29 in 947772d

   class Table(FilterElement, component='table.js'):

3. Buckaroo can currently send a complete dataframe via JSON, but not parquet. I probably want to enable sending a complete dataframe via parquet - parquet is much faster and has more expressive types that are closer to python/dataframe types.
4. I'm not sure how to adapt infinite row scrolling to the nicegui model. This should be possible but would require discussion.

What do you all think?

[falkoschindler]

Hi @paddymul,

Thanks for sharing Buckaroo and showing how it might connect with NiceGUI.

First impression

Buckaroo’s data-grid features look interesting and could add functionality that some users have asked for in the past, but we haven’t yet evaluated it in depth.

Next steps

Our bandwidth is limited right now, so the best way forward would be a small proof-of-concept led by you or other community members. Ideally it would:

• take a pandas DataFrame (or similar) as input,
• render through Buckaroo inside a NiceGUI page,
• expose just a minimal, well-documented set of options,
• keep the additional frontend bundle reasonably small.

If such a demo appears, we can review it and decide together whether Buckaroo should remain an external plugin (nicegui-buckaroo) or could later move into core.

Community members: if this integration would help your projects, please upvote, add a 👍 or share your use-case—those signals will help us prioritise.

Thanks again for the proposal, and looking forward to any prototypes the community puts together!