🪶 liteprofile

A lightweight, fast alternative inspired by ydata-profiling (pandas-profiling) — built on Polars.

⚡️ Why

liteprofile is inspired by the great work of ydata-profiling.
While that library provides rich and detailed statistical reports, liteprofile focuses on speed, simplicity, and small output for quick exploratory data analysis (EDA).

It’s designed as a complement — use ydata-profiling when you need deep insights, and liteprofile when you need a fast, clear overview.

🚀 Features

• ⚡ Super-fast summaries with Polars or DuckDB
• 📊 Numeric stats: mean, std, quantiles, outliers
• 🔢 Categorical summaries: top frequencies
• 🔗 Optional numeric correlations
• 🧠 Smart warnings (constant / missing / high-cardinality)
• 🧰 CLI and Python API
• 💾 Outputs: Markdown or minimal HTML
• 🧩 Easy to extend — build your own “lite” analytics blocks

🧰 Installation

pip install liteprofile
# or from source
pip install git+https://github.com/inezvl/liteprofile.git

💡 Quickstart

From Python

import polars as pl
from liteprofile import profile, profile_html

df = pl.DataFrame({
    "id": [1, 2, 3, 4, 5],
    "city": ["Antwerp", "Ghent", "Ghent", "Brussels", None],
    "price": [100.0, 200.5, 180.2, 300.0, 300.0]
})

# Markdown summary
print(profile(df))

# HTML summary
html = profile_html(df)
with open("report.html", "w", encoding="utf-8") as f:
    f.write(html)

From CLI

python -m liteprofile data.csv --html --out report.html
open report.html

🪶 Example Output (Markdown)

# liteprofile report

## Overview
| metric          | value |
|-----------------|--------|
| rows            | 5      |
| columns         | 3      |
| duplicate_rows  | 0      |

## Columns
| column | dtype | nulls | null_% | unique | mean | std | min | q1 | median | q3 | max | outliers_iqr | sample |
|--------|--------|-------|--------|---------|------|-----|-----|----|---------|----|-----|---------------|---------|
| id     | Int64  | 0 | 0 | 5 | 3 | 1.58 | 1 | 2 | 3 | 4 | 5 | 0 | [1, 2, 3, 4, 5] |
| city   | Utf8   | 1 | 20 | 3 |  |  |  |  |  |  |  |  | ['Antwerp', 'Ghent', 'Ghent', 'Brussels', None] |
| price  | Float64 | 0 | 0 | 3 | 216 | 83.8 | 100 | 180 | 200 | 300 | 300 | 0 | [100.0, 200.5, 180.2, 300.0, 300.0] |

🧩 Roadmap

🧠 Philosophy

liteprofile aims to complement existing data-profiling tools.
Instead of competing, it offers a minimal mode for everyday use — perfect for quick checks before going deeper with heavier frameworks.

Inspired by the community feedback around the need for a “fast mode” in ydata-profiling.

🤝 Contributing

Contributions are welcome!
You can help by:

• Adding backends (DuckDB, Arrow, SQLite)
• Enhancing HTML rendering
• Building small extensions (e.g., histograms, missing-value heatmaps)
• Benchmarking against other profilers

📜 License

MIT © Inez Van Laer

⭐️ Support

If you like the project:

• Leave a ⭐️ on GitHub — it really helps
• Open a Discussion for feedback or feature ideas
• Tell us what you’d love to see in the next release!