an anywidget for data that talks like a duck
Warning
Here be dragons 🐉. quak is a prototype and not production-ready. Expect bugs.
The library is open-sourced for SciPy 2024.
pip install quakThe easiest way to get started with quak is using the IPython cell magic.
%load_ext quakimport pandas as pd
df = pd.read_csv("https://raw.githubusercontent.com/vega/vega-datasets/main/data/airports.csv")
dfAny cell that returns an object implementing the
Python dataframe interchange protocol
(i.e., a dataframe-like "thing") will be rendered using quak.Widget, rather
than the default renderer.
Alternatively, you can use quak.Widget directly:
import polars as pl
import quak
df = pl.read_csv("https://raw.githubusercontent.com/vega/vega-datasets/main/data/airports.csv")
quak.Widget(df)quak requires both rye (for Python) and deno (for TypeScript).
If you want to develop in the notebooks (./examples/), you will need to run
both deno (to (re)build the TypeScript) and rye (to start the Jupyter
notebook):
deno task devand then start the Python notebook server with rye:
rye sync
rye run jupyter labAlternatively, you can just work on the TypeScript side of things by running:
npx vite
# or deno run -A npm:viteand editing ./lib/example.ts.
We check linting and formatting in CI:
# typescript
deno lint
deno fmt
deno task check
# python
rye lint
rye formatNote
Why the weird TypeScript stuff? In practice, hybrid Python/JS repos get messy
with npm and node_modules. With Deno there is no node_modules, and the
tool handles type-checking, linting, and formatting. The extra build scripts
serve to make nice development ergonamics within and outside of Jupyter.