1.0.1

[1.0.1] - 2025-09

Minor bug fixes and updates.

[1.0] - 2025-09

New major version of feets, with a full redesign and modernization.

Support for older versions of Python has been removed. feets now officially supports Python 3.10+.

Highlights of this release

Parallelized workflow & batch processing

A new Dask-powered parallelized model has been introduced for feature extraction.

• Dramatically reduces runtime for large datasets.
• Enables scalable batch processing of multiple light curves simultaneously using the new FeatureSpace.extract_many() method.
• Simplifies large-scale analyses and integration with machine learning pipelines by processing multiple datasets in a single call.

Integration with light-curve

This release incorporates all extractors available in the light-curve library.

• Existing extractors in feets have been replaced by optimized implementations from light-curve, improving speed, numerical stability, and reliability.
• Additional extractors from light-curve have been added, expanding the feature set.
• New compatibility with flux and flux_error data vectors, allowing extractors to operate directly on flux-based inputs as well as magnitudes.

Configuration persistence

New functionality for saving and loading FeatureSpace configurations has been introduced.

• Configurations can now be exported and imported in JSON and YAML formats.
• Methods FeatureSpace.to_json() and FeatureSpace.to_yaml(), as well as functions feets.read_json() and feets.read_yaml(), enable reproducible workflows and easy sharing of feature extraction setups.
• This makes it straightforward to store, reuse, and distribute predefined extraction pipelines across projects.

Main changes to the API

feets.Extractor

The extractor system has been fully redesigned for clarity, configurability, and performance.

• Removed: the data, dependencies, and params attributes.
• Changed: the fit() method has been renamed to extract().
• Changed: data requirements and feature dependencies are now inferred directly from the signature of the extract() method.
• Changed: extractor parameters are now defined in the __init__() constructor instead of a params dictionary.
• Added: flatten_feature() method for custom flattening of multi-value features.
• Added: compatibility with new data vectors (flux, flux_error).

feets.FeatureSpace

• Changed: the extract() method now accepts light-curve data as keyword arguments (e.g., extract(time=t, magnitude=m)).
• Added: extract_many() method for batch feature extraction from multiple light curves.
• Added: from_dict(), to_dict(), to_json(), and to_yaml() methods for configuration persistence and reproducibility.
• Added: from_lightcurve() and from_lightcurves() for automatic feature selection based on the available data.

feets.features

• Added: Features class for a structured representation of extracted features. Encapsulates feature values, names, and associated metadata in a single object, with support for easy conversion to dictionaries, pandas.DataFrames, and other formats for downstream analysis.

feets.registry

• Added: ExtractorRegistry class for registration, discovery, and ordering of extractors.
• Added: a central extractor_registry object that stores available extractors and features, ensuring correct execution order and dependency resolution.

feets.io

• Added: load_json() and load_yaml() functions for reading existing FeatureSpace configurations.

feets.runner

• Added: run() function for running instances of feature extractors in parallel with dask.