Skip to content

Product Format Plugins

The SCT framework is designed to be analysis-agnostic with respect to input product formats.
Different satellite and sensors products may have varying data formats and conventions.
To decouple the analysis code from the input data type, we use a plugin-based architecture.

This allows:

  • New input product formats to be added without modifying the core analysis code.
  • Analyses to interact with a unified interface, regardless of the actual product type.
  • Easy distribution of input product plugins as independent packages.

Several plugins for some of the most common SAR products and sensors are already available.

Supported Products

To use the functionalities of SCT for a specific product, it is necessary to install, in addition to SCT, the corresponding plugin. Plugins are Python packages that can be discovered automatically by SCT once installed.

Architecture

The plugin system has been implemented with the following architecture:

  1. Protocol Layer (sct.plugins.input_products_protocols)

    • Defines the protocols (interfaces) that all input product plugins must implement.
    • Contains core abstractions such as InputProductPluginProtocol and AbsoluteLocalizationErrorCorrector.
    • No knowledge of how plugins are loaded or installed.

  2. Loader Layer (sct.plugins.loader)

    • Responsible for discovering and loading plugins.
    • Uses stevedore to dynamically load all installed plugins in the sct.input_products namespace.
    • Isolated from analysis logic.

  3. Registry Layer (sct.plugins.__init__)

    • Provides a registry of loaded plugins.
    • Analysis code interacts only with the registry, never with plugin loading mechanisms.

  4. Plugin Packages

    • Each plugin lives in a separate installable package.
    • Declares an entry point in pyproject.toml using the sct.input_products namespace.
    • Implements the InputProductPluginProtocol protocol.

Plugin Discovery with Stevedore

Plugins are discovered dynamically using Stevedore:

Discovery Mechanism
from stevedore import ExtensionManager

manager = ExtensionManager(
    namespace="sct.input_products",
    invoke_on_load=True,
    on_load_failure_callback=_on_load_failure
)

Namespace must match the entry point group defined in each plugin's pyproject.toml.

Benefits

  • Extensibility: New input products are supported via independent packages.
  • Decoupling: Core analysis logic does not depend on plugin implementation.
  • Testability: Plugins can be mocked or substituted without touching the core.
  • Dynamic discovery: Plugins are discovered at runtime via entry points, no hard-coded imports required.

Installation and Usage

To install a specific plugin for SCT using pip:

Installing a Plugin
pip install <name_of_sct_plugin>

Then the software can discover automatically all the installed plugins and perform analyses on products corresponding to the installed plugins.

Product Format Plugins Documentation

A dedicated documentation site for SCT Input Product Plugins has been created to gather all plugin python packages developed and all the supported sensors and formats.

Refer to the official plugins documentation for further information.