analysta 0.0.4

DataFrame Toolkit for Analysts

analysta 🖇️

A Python library for comparing pandas DataFrames using primary keys, tolerances, and audit-friendly diffs.** Easily detect mismatches, missing rows, and cell-level changes between two datasets.

---

🧾 Table of Contents

• Installation
• Quick Example
• More Examples
• Features
• License

🚀 Installation

pip install analysta

Python 3.8 or higher is required.

For a concise import, you can alias the package:

import analysta as nl

⚡ Quick Example

import analysta as nl
import pandas as pd

# Row 1 exists only in df1, row 4 only in df2
# Row 3 exists in both but has a different price
df1 = pd.DataFrame({"id": [1, 2, 3], "price": [100, 200, 300]})
df2 = pd.DataFrame({"id": [2, 3, 4], "price": [200, 250, 400]})

delta = nl.Delta(df1, df2, keys="id")
print(delta.unmatched_a)         # → id=1
print(delta.unmatched_b)         # → id=4
print(delta.changed("price"))    # → id=3
print(nl.find_duplicates(df1, column="id"))  # Duplicates by column

📚 More Examples

Tolerant numeric diffs

import analysta as nl
import pandas as pd

df_a = pd.DataFrame({"id": [1, 2], "value": [100.0, 200.005]})
df_b = pd.DataFrame({"id": [1, 2], "value": [100.0, 200.0]})

delta = nl.Delta(df_a, df_b, keys="id", abs_tol=0.01)
print(delta.changed("value"))  # diff 0.005 < 0.01 → empty

delta = nl.Delta(df_a, df_b, keys="id", abs_tol=0.001)
print(delta.changed("value"))  # diff 0.005 > 0.001 → id=2

Counting duplicates

df = pd.DataFrame({"id": [1, 1, 2, 2, 2]})
print(nl.find_duplicates(df, column="id", counts=True))

Trimming whitespace

df = pd.DataFrame({"id": ["1"], "name": [" Alice "]})
clean = nl.trim_whitespace(df)
print(clean)

✨ Features

• Key-based row comparison: "A not in B" and vice versa
• Tolerant numeric diffs (absolute & relative)
• Highlight changed columns
• Built for analysts, not just engineers
• Automatic trimming of leading/trailing whitespace
• Detect duplicate rows with optional counts
• CLI and HTML reporting coming soon

📄 License

analysta is distributed under the terms of the MIT license.