Lontras

⚠️ This library is under development and has not been released yet

We love pandas and its siblings! They're the industry-standard tools for powerful data manipulation. However, for smaller projects or when minimizing dependencies is paramount, lontras offers a lightweight, pure-Python alternative built on simple dictionaries. Designed for ease of use and direct integration, lontras encourages you to copy and paste its core components into your code. It's the perfect stand mixer for baking smaller data tasks when bringing out the full industrial equipment of pandas is too much.

TLDR;

• 🤏 Small-scale DataFrame operations? lontras!
• 🚀 Fast DataFrame loading, no dependencies? lontras!
• ⚙️ Embed DataFrames in MicroPython? lontras!
• 🌐 Use DataFrames in the browser (PyScript)? lontras!
• 🤝 High Pandas compatibility? lontras!
• 📦 1000x smaller! Pandas+Numpy ~ 120MB; lontras ~ 102KB!
• ⏱️ Loads 20x faster! Pandas ~ 400ms; lontras ~ 20ms! test script

Table of Contents

• Installation
• Documentation
• Core Functionalities
• License

Installation

If you prefer to install lontras using pip for easy management and updates, you can do so with the following command:

pip install lontras
# or
uv pip install lontras

This will download and install lontras from the Python Package Index (PyPI).

For lightweight projects where you want to avoid external dependencies, you can simply copy the source file src/lontras/lontras.py into your project directory. This allows you to directly use the library functions from your code without any installation.

Documentation

Check out the API Docs.

Core Functionalities:

Lontras prioritizes simplicity and minimal dependencies. It leverages Python's native dictionaries (via UserDict) to offer core DataFrame and Series functionalities without external libraries.

Data Structures:

• Series: A one-dimensional array-like structure.
• DataFrame: A two-dimensional labeled data structure.

Accessing Data:

• loc and iloc: Access data by label or by index.
• Label-based access: Access data using standard dictionary-like syntax (e.g., series['label']).
• Positional access (slicing): Use slices for location-based access (e.g., series[1:3]).
• Attribute-based access: For convenient access to all values for a given key, use attribute-based access using getattr that dynamically retrieves data based on provided keys or a list of keys.

Modifying Data:

• Setting values: Modify existing values or add new entries using loc or iloc assignment (e.g., series.loc['label'] = value).
• Deleting values: Remove entries using del series['label'].
• Concatenation: Combine Series or DataFrames vertically or horizontally.

Transforming Data:

• Mapping and applying functions: Apply functions element-wise using map or along axes/indices using apply.
• Sorting: Sort indexes and values using provided sorting functions.
• Basic operations: Use standard Python operators (+, -, *, /, //, %, **, comparisons) for element-wise operations.

Data Aggregation and Combination:

• groupby: Group data based on a column and perform operations within each group (similar to pandas groupby).
• Join/Merge: Merge two DataFrames based on specific columns (similar to pandas join/merge operations).
• reduce: Apply a function cumulatively to the elements. Basic reduction functions like max, min, argmax, sum, etc.
• Leverages Python's built-in statistics module for basic statistical calculations.

Limitations & Trade-offs: (Same as before)

• No dtype! You ask for a sum, lontras will try to sum and may raise an exception if an unexpcted value is found.
• Specialized Data Handling: Lontras focuses on core functionalities and doesn't include specialized functions for datetime, strings, or categorical data. However, users can achieve similar behavior through apply and map functions.
• Statistical Functions: Limited set of statistical functions. Lontras relies primarily on Python's built-in statistics module.
• Data Import/Export: Supports limited import/export formats. External libraries might be necessary for complex file handling.
• Missing Data Handling: Doesn't include dedicated functions for handling missing data. Users can implement their own logic using conditional statements or filtering.
• Multilevel Indexing: Lacks built-in support for multilevel indexing. However, tuple indexes can be used to achieve similar hierarchical structures.
• Plotting: Currently doesn't include plotting functionalities. External plotting libraries are recommended for visualization.

License

lontras is distributed under the terms of the MIT license.