ipy-compare

ipy-compare is an interactive tool designed for use in Jupyter notebooks and Google Colab, enabling users to visually compare rows and columns of a DataFrame. Ideally used for tagging or for labeling data in a pandas DataFrame.

Features

• Row and Column Comparison: Supports both overall row-level measures and column-specific measures.
• Pagination: Navigate through rows using custom indices or iterators.
• Repeatable Sampling: Use a fixed random seed for consistent sampling.
• Branding: Add customizable branding to the footer, including a linkable logo and text.
• Interactive Radio Buttons: Visual indicators for selected measures.
• Colab Compatible: Works seamlessly in Google Colab notebooks.

Example:

We support two different ways to use ipy-compare.

1. Each Column: Use a measure for each column.
2. Overall Row: Use a measure for the entire row.

Where you can specify the questions you want to ask for each column or for the entire row.

Each Column

Overall Row

Installation

pip install ipy-compare

Usage

Import and Initialize

from ipy_compare import Compare
import pandas as pd
import panel as pn  # Required for notebook/colab display

# Sample DataFrame
data = {
    'Column1': ['A', 'B', 'C'],
    'Column2': ['X', 'Y', 'Z'],
    'Column3': ['Apple', 'Fish swimming', 'Vrooom car']
}
df = pd.DataFrame(data)

# Define measures
measures = {
    "overall": ["Good", "Bad"],
    "each": ["Better", "Worse", "Neutral"]
}

# Initialize Compare
tool = Compare(df, columns=["Column1", "Column3"], measures=measures)

# Display the tool
tool.servable()

Navigation and Interaction

1. Navigate between rows using the Previous and Submit & Next buttons.
2. Use the radio buttons to select measures for each column or for the overall row.
3. Submit measurements using either the Submit button or Submit & Next to move to the next row.
4. All measurements are captured and can be retrieved programmatically.

Custom Pagination

To specify a subset of rows or custom order:

# Custom row order
custom_order = [2, 0, 1]

# Initialize Compare with custom pagination
tool = Compare(df, columns=["Column1", "Column3"], measures=measures, pagination=custom_order)
tool.servable()

Repeatable Sampling

To use random sampling with a fixed seed:

# Get a random sample of rows
sampled_indices = Compare.sample_indices(df, n=2, seed=42)

# Initialize Compare with sampled indices
tool = Compare(df, columns=["Column1", "Column3"], measures=measures, pagination=sampled_indices)
tool.servable()

Retrieve Measurements

Once interactions are complete, retrieve the captured measurements:

# Get measurements as a DataFrame
measurements = tool.get_measurements()
print(measurements)

Example Output:

Google Colab Usage

When using in Google Colab, make sure to run these commands first:

!pip install ipy-compare panel

The tool is fully compatible with Google Colab and will work the same way as in Jupyter notebooks.

Advanced Options

Custom Measures

Define your own measure categories:

measures = {
    "overall": ["Excellent", "Poor"],
    "each": ["Correct", "Incorrect", "Not Sure"]
}

Contributing

Contributions are welcome! Please fork the repository and submit a pull request with your changes.

License

This project is licensed under the MIT License.