Metadata-Version: 2.4
Name: bea-tools
Version: 0.0.6
Summary: Tools made by, and for, Bea - a Python toolkit for data sampling and analysis workflows
License: MIT
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering
Requires-Python: >=3.10
Description-Content-Type: text/markdown
Requires-Dist: pandas>=2.2
Requires-Dist: pydicom>=3.0

# bea-tools

**🐝 𝓉𝑜𝑜𝓁𝓈 𝓂𝒶𝒹𝑒 𝒷𝓎, 𝒶𝓃𝒹 𝒻𝑜𝓇, 𝒷𝑒𝒶 🐝**

A Python package of random functions and tools that I use regularly. Data science/analysis focused since I'm a data scientist c:

## Installation

```bash
pip install bea-tools
```

## Features

### TreeSampler

A hierarchical stratified sampling tool for pandas DataFrames. Designed for scenarios where you need to sample data while maintaining specific proportions across multiple categorical dimensions, with intelligent handling of capacity constraints.

**Key capabilities:**

- **Multi-dimensional stratification**: Define sampling targets across multiple features (e.g., gender, age group, category)
- **Hierarchical spillover**: When a stratum lacks sufficient data, excess quota automatically redistributes to sibling strata
- **Flexible matching**: Match values using `equals`, `contains`, or `between` strategies
- **Conditional weights**: Define weights that vary based on the path through the sampling tree
- **Strict mode**: Lock specific strata to prevent them from absorbing spillover
- **Single-per-entity sampling**: Ensure unique entities (e.g., one exam per patient)

### DicomComparer

A tool for comparing two DICOM files at the attribute level. Identifies which attributes are shared between files, which are exclusive to each, and whether shared attributes have matching or conflicting values.

**Key capabilities:**

- **Attribute overlap analysis**: Identify which DICOM tags exist in both files vs. exclusive to one
- **Value comparison**: For shared attributes, detect matches and conflicts
- **Flexible input**: Pass files directly or as a labeled dictionary
- **Summary output**: Generate formatted comparison reports

```python
import pydicom
from bea_tools._dicom.dicomp import DicomComparer

dcm1 = pydicom.dcmread("path/to/first.dcm")
dcm2 = pydicom.dcmread("path/to/second.dcm")

comparer = DicomComparer(dcm1, dcm2)
comparison = comparer.compare()

# Print summary statistics
comparison.summary()

# Access specific conflicts
for conflict in comparison.intersection.comparison.conflicts:
    print(conflict)
```

## Quick Start

```python
from bea_tools import TreeSampler
from bea_tools._pandas.sampler import Feature

import pandas as pd

# Sample data
df = pd.DataFrame({
    'patient_id': ['P001', 'P002', 'P003', 'P004', 'P005', 'P006'],
    'gender': ['M', 'M', 'F', 'F', 'M', 'F'],
    'age': [25, 45, 35, 55, 30, 40],
    'studydate_anon': pd.date_range('2020-01-01', periods=6)
})

# Define stratification features
features = [
    Feature(
        name='gender',
        match_type='equals',
        levels=['M', 'F'],
        weights=[0.5, 0.5]  # Target 50/50 split
    )
]

# Create sampler and extract stratified sample
sampler = TreeSampler(
    n=4,                          # Target sample size
    features=features,
    seed=42,                      # For reproducibility
    count_col='patient_id',       # Column for unique entity identification
    single_per_patient=True       # One row per patient
)

result = sampler.sample_data(df)
```

## Advanced Usage

### Age Brackets with Between Matching

```python
age_feature = Feature(
    name='age',
    match_type='between',
    levels=[(0, 30), (30, 50), (50, 100)],
    weights=[0.3, 0.4, 0.3],
    labels=['Young', 'Middle', 'Senior'],
    label_col='age_group'
)
```

### Strict Strata (No Spillover)

```python
# This stratum will maintain exact proportions, never absorbing excess
feature = Feature(
    name='category',
    match_type='equals',
    levels=['A', 'B'],
    weights=[0.7, 0.3],
    strict=True  # Prevents spillover absorption
)
```

### Conditional Weights

Define weights that depend on parent feature values:

```python
category_feature = Feature(
    name='category',
    match_type='equals',
    levels=['X', 'Y'],
    conditional_weights=[{
        'feature': 'gender',
        'weights': {
            'M': [0.6, 0.4],  # When gender=M: 60% X, 40% Y
            'F': [0.4, 0.6]   # When gender=F: 40% X, 60% Y
        }
    }]
)
```

## Requirements

- Python 3.10+
- pandas >= 2.2
- pydicom (for DICOM utilities)

## License

MIT
