mitoric 0.0.1.dev3

Polars-native EDA and comparison report generator that outputs a single self-contained HTML.

日本語版: README_JP.md

mitoric

A library that generates EDA / comparison reports for Polars DataFrames as a single HTML file.

PyPI

Features

• Generate a report for a single dataset
• Generate a comparison report for two datasets
• Dataset summary (rows/columns/missing/duplicates, etc.)
• Column profiles (numeric/categorical/text/datetime/boolean)
• Correlation/association calculations
• Histogram bin size switching (10/15/30/50)

Requirements

• Python 3.10+
• polars>=1
• jinja2>=3.1.6

Usage

Report for a single dataset

import polars as pl
from mitoric import generate_single_report

frame = pl.DataFrame(
    {
        "age": [10, 12, 12, 14],
        "city": ["A", "B", "A", "C"],
    }
)

html = generate_single_report(frame, save_path="/tmp/report.html")
print(html[:200])

Comparison report for two datasets

import polars as pl
from mitoric import generate_compare_report

left = pl.DataFrame({"value": [1, 2, 3]})
right = pl.DataFrame({"value": [2, 3, 4]})

html = generate_compare_report(
    left,
    right,
    left_name="before",
    right_name="after",
    save_path="/tmp/compare_report.html",
)

Explicit column type specification

import polars as pl
from mitoric import generate_single_report
from mitoric.models.base import ColumnName, ColumnType, ExplicitType

frame = pl.DataFrame({"flag": ["yes", "no", "yes"]})
explicit_types = [ExplicitType(ColumnName("flag"), ColumnType("categorical"))]

html = generate_single_report(frame, explicit_types=explicit_types)

API

• generate_single_report(frame, *, target_columns=None, explicit_types=None, save_path=None)
• generate_compare_report(left, right, *, target_columns=None, explicit_types=None, save_path=None, left_name=None, right_name=None)

Types supported in explicit_types: numeric, categorical, text, datetime, boolean

Examples

• examples/output/single_report.html
• examples/output/compare_report.html

Constraints and Notes

• Only Polars is supported (Pandas is not supported)
• Generated HTML depends on TailwindCSS and chart.js via CDN
• If save_path is not specified, HTML is returned as a string and nothing is saved

Disclaimer

This library was built with a fairly high ratio of vibe coding, so it may include unnecessary processing, questionable design, or tests. We will improve it as we find issues.

For Developers

Setup

uv sync

Development

uv sync
make lint
make test
make test_e2e # E2E outputs are saved under examples/output/ for regression checks