https://github.com/ivankmk/dfaudit
Audit your pandas DataFrame before you trust it - missing values, soft missing, cardinality and top categories in one call.
https://github.com/ivankmk/dfaudit
data-audit data-governance data-profiling data-quality eda exploratory-data-analysis matplotlib missing-data pandas
Last synced: 10 days ago
JSON representation
Audit your pandas DataFrame before you trust it - missing values, soft missing, cardinality and top categories in one call.
- Host: GitHub
- URL: https://github.com/ivankmk/dfaudit
- Owner: ivankmk
- License: mit
- Created: 2026-05-05T17:21:29.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2026-05-10T14:31:52.000Z (about 1 month ago)
- Last Synced: 2026-05-10T16:29:44.838Z (about 1 month ago)
- Topics: data-audit, data-governance, data-profiling, data-quality, eda, exploratory-data-analysis, matplotlib, missing-data, pandas
- Language: Python
- Homepage:
- Size: 516 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
dfaudit
Know if you can trust your data before you use it.
A fast, modern data audit tool for pandas DataFrames -- built for analysts who need to assess data reliability, not just explore it.
[](https://pypi.org/project/dfaudit/)
[](https://www.python.org/)
[](LICENSE)
---
## What is dfaudit?
**dfaudit** is a lightweight Python package for auditing pandas DataFrames. Before you clean, model, or report -- you need to know whether your data is reliable. dfaudit gives you that answer fast.
Think of it as an auditor's first pass: one call to understand how complete, consistent, and trustworthy your data actually is. Missing values, cardinality, dominant categories -- the things that tell you if the data holds up.
## Install
```bash
pip install dfaudit
```
## Quickstart
```python
import pandas as pd
import dfaudit as dfa
df = pd.read_csv("https://raw.githubusercontent.com/pandas-dev/pandas/master/doc/data/titanic.csv")
dfa.overview(df)
dfa.missing_matrix(df)
dfa.unique_matrix(df)
```
## What you get
### `overview(df)` - Column-level data quality table
A styled summary table covering every column at once: dtypes, missing counts and percentages, unique value counts, mode, and top-3 categories - all color coded so problems jump out instantly.
**missing** — standard `NaN` / `None` / `NaT` values pandas considers null.
**soft_missing** — values that are technically non-null but meaningless in practice: empty strings, placeholder text, and numeric sentinel codes commonly used in legacy systems or exports. dfaudit flags these automatically:
- Strings: `""`, `"n/a"`, `"na"`, `"null"`, `"none"`, `"nan"`, `"nil"`, `"-"`, `"--"`, `"?"`, `"unknown"`, `"missing"`, `"not available"`, `"not applicable"`
- Numbers: `-999`, `-9999`, `9999`, `99999`
Missing percentage cells are highlighted red for high missingness, blue for high cardinality.
### `missing_matrix(df)` - Row-level missingness visualization
A matrix plot that shows *where* your data is missing across every row, with a bar chart of missing percentages per column on the right. Spot correlated missingness, systematic gaps, and data loading issues that column-level stats miss entirely.
```python
dfa.missing_matrix(df, style="vivid")
```
### `unique_matrix(df)` - Row-level unique value visualization
A matrix plot that shows the distribution of unique values across every row and column. Low-cardinality columns (≤ 10 unique values) get a distinct color per value — so a binary column like `Sex` renders in two colors, a 3-class column like `Pclass` in three. High-cardinality columns (`Name`, `Ticket`, `PassengerId`) appear in a uniform gray, making it immediately obvious which columns are categorical vs. continuous or near-unique identifiers. A bar chart on the right shows the exact unique value count per column.
```python
dfa.unique_matrix(df)
dfa.unique_matrix(df, style="vivid", max_colors=8)
```
## Why I built this
I work with a lot of different data on a daily basis. For years I carried the same audit snippets from notebook to notebook -- copy, paste, tweak, repeat. At some point I decided to stop doing that and package the most common data audit tasks into something reusable, with visuals I actually want to look at.
dfaudit is what I reach for now at the start of every new dataset. I hope it saves you the same time it saves me.
If you work with modern data and want to contribute -- please do, pull requests are very welcome.
---