additory 0.1.0a4

A semantic, extensible dataframe transformation engine with expressions, lookup, and synthetic data generation support.

Additory

A semantic, extensible dataframe transformation engine with expressions, lookup, and augmentation support.

Author: Krishnamoorthy Sankaran

🛠️ Requirements

• Python: 3.9+
• Core dependencies: pandas, polars, numpy, scipy
• Optional: cuDF (for GPU support)

📦 Installation

pip install additory==0.1.0a4

Optional GPU support:

pip install additory[gpu]==0.1.0a4  # Includes cuDF for GPU acceleration

Development installation:

pip install additory[dev]==0.1.0a4  # Includes testing and development tools

🎯 Core Functions

Function	Purpose	Example
add.to()	Lookup/join operations	add.to(df1, from_df=df2, bring='col', against='key')
add.synthetic()	Generate additional data	add.synthetic(df, n_rows=1000)
add.deduce()	Text-based label deduction	add.deduce(df, from_column='text', to_column='label')
add.scan()	Data profiling & analysis	add.scan(df, preset="full")

🧬 Available Expressions

Additory includes 12 built-in health and fitness expressions:

• add.bmi() - Body Mass Index
• add.bsa() - Body Surface Area
• add.bmr() - Basal Metabolic Rate
• add.waist_hip_ratio() - Waist-to-Hip Ratio
• add.body_fat_percentage() - Body Fat Percentage
• add.ideal_body_weight() - Ideal Body Weight
• add.blood_pressure_category() - BP Classification
• add.cholesterol_ratio() - Cholesterol Ratio
• add.age_category() - Age Classification
• add.fitness_score() - Overall Fitness Score

# Health calculations
patients = pd.DataFrame({
    'weight_kg': [70, 80, 65],  # Weight in kilograms
    'height_m': [1.75, 1.80, 1.60],  # Height in meters
    'age': [25, 35, 45],
    'gender': ['M', 'F', 'M']
})

patients_bmi = add.bmi(patients)
patients_bsa = add.bsa(patients)
fitness_scores = add.fitness_score(patients)

# Chain multiple expressions
result = add.fitness_score(add.bmr(add.bmi(patients)))

🔧 DataFrame Support

Additory works seamlessly with multiple DataFrame libraries:

• pandas - Full support
• polars - Full support
• cuDF - GPU acceleration support

import polars as pl
import additory as add

# Works with polars
df_polars = pl.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
result = add.synthetic(df_polars, n_rows=100)

# Automatic type detection and conversion

✨ Key Features

🔧 Utilities

add.to() - Data Lookup & Joins Simplified syntax for bringing columns from one dataframe to another.

# Simple lookup
orders_with_prices = add.to(
    orders, 
    from_df=products, 
    bring='price', 
    against='product_id'
)

# Multiple columns and keys
enriched = add.to(
    orders,
    from_df=products,
    bring=['price', 'category'],
    against=['product_id', 'region']
)

add.onehotencoding() - Categorical Encoding Convert categorical columns to one-hot encoded format.

# One-hot encoding (single column)
encoded = add.onehotencoding(df, 'category')

add.harmonize_units() - Unit Standardization Standardize units across your dataset.

# Unit harmonization
standardized = add.harmonize_units(
    df, 
    value_column='temperature', 
    unit_column='unit',
    target_unit='C'
)

🧮 Expressions

Pre-built calculations for health, fitness, and common metrics. Simple examples:

# Create patient data with correct column names
patients = pd.DataFrame({
    'weight_kg': [70, 80, 65],  # Weight in kilograms
    'height_m': [1.75, 1.80, 1.60],  # Height in meters
    'age': [25, 35, 45],
    'gender': ['M', 'F', 'M']
})

# Calculate BMI
patients_with_bmi = add.bmi(patients)

# Calculate Body Surface Area
patients_with_bsa = add.bsa(patients)

# Chain multiple expressions
result = add.fitness_score(add.bmr(add.bmi(patients)))

🔄 Synthetic Data Generation

Synthetic generates additional data similar to your existing dataset using inline strategies.

# Extend existing data (learns from patterns)
more_customers = add.synthetic(customers, n_rows=1000)

# Create data from scratch with strategies
new_data = add.synthetic("@new", n_rows=500, strategy={
    'id': 'increment:start=1',
    'name': 'choice:[John,Jane,Bob]',
    'age': 'range:18-65'
})

🤖 Text-Based Label Deduction

Deduce automatically fills in missing labels by learning from your existing labeled examples. Pure Python, no LLMs, offline-first.

# Deduce missing labels from text
tickets = pd.DataFrame({
    "ticket_text": ["Cannot log in", "Billing question", "App crashes", "Need invoice"],
    "category": ["Technical", "Billing", None, None]
})

# Automatically fill in missing categories
result = add.deduce(tickets, from_column="ticket_text", to_column="category")

# Use multiple columns for better accuracy
result = add.deduce(
    df,
    from_column=["title", "description"],
    to_column="category"
)

🧪 Examples

E-commerce Data Pipeline

import pandas as pd
import additory as add

# Start with small customer sample
customers = pd.DataFrame({
    'customer_id': [1, 2, 3],
    'age': [25, 35, 45],
    'region': ['North', 'South', 'East']
})

# Generate more customers
customers = add.synthetic(customers, n_rows=10000)

# Add customer tiers
tiers = pd.DataFrame({
    'customer_id': range(1, 4),  # Match original IDs
    'tier': ['Gold', 'Silver', 'Bronze']
})

# Use pipeline approach
result = (customers
    .pipe(add.to, from_df=tiers, bring='tier', against='customer_id')
    .pipe(add.scan, preset="quick"))

print(result.summary())

Healthcare Data Analysis

# Create patient data from scratch
strategy = {
    'patient_id': 'increment:start=1',
    'age': 'range:18-80',
    'weight_kg': 'range:50-120',  # Weight in kg
    'height_cm': 'range:150-200'  # Height in cm
}

patients = add.synthetic("@new", n_rows=1000, strategy=strategy)

# Convert height to meters for expressions
patients['height_m'] = patients['height_cm'] / 100

# Calculate health metrics using pipeline
result = (patients
    .pipe(add.bmi)
    .pipe(add.scan, preset="correlations"))

print(result.correlations)

📚 Documentation

• Function Documentation - Detailed guides for each function
• Expressions Guide - Complete expressions reference

📄 License

MIT License - see LICENSE file for details.

📞 Support

• Issues: GitHub Issues
• Documentation: Full Documentation

🗺️ v0.1.1 (January 2026)

• Enhanced documentation and tutorials
• Performance optimizations
• Additional expressions
• Advanced synthetic data patterns

---

Made with ❤️ for data scientists, analysts, and developers who love working with data.