Prototyping dataframe visualization tools

This package exists to experiment with ideas and features that could be included in skrub. As the goal is quick prototyping and experimentation, some features are likely to be incomplete or dysfunctional.

See some example reports. Any feedback is very welcome.

Installing

skrubview is not available on PyPI nor conda-forge. It must be installed by downloading the latest source code. Either in one go:

pip install git+https://github.com/skrub-data/skrubview.git

Or by cloning the repository:

git clone git@github.com:skrub-data/skrubview.git
pip install ./skrubview

Usage

from sklearn.datasets import load_iris
from skrubview import Report

df = load_iris(as_frame=True)["frame"]
Report(df).open_html()

Note: the screenshot above is outdated.

Reports can be initialized with either a dataframe (pandas or polars) or a file path. They should display nicely in a jupyter notebook or ipython shell. They have html, text, and json attributes representing their content in those formats.

Command-line usage

skrubview ./my_file.parquet --open

See skrubview -h for options.