⚡ [Feature]: Implement a Centralized Configuration System

[PPeitsch]

Problem Description

Currently, algorithm parameters (e.g., smoothing_window, peak_prominence, min_points_fit) are hardcoded as default values in the signatures of many different functions and methods across the codebase. This approach has several drawbacks:

1. Inconsistency: It's difficult to ensure that related parameters are consistent across different modules.
2. Difficult to Tune: To change a default behavior globally, a developer must find and modify every instance of that parameter, which is error-prone.
3. Reproducibility Issues: If a default parameter value changes between library versions, users' scripts might produce different results without any code changes, harming reproducibility.

Proposed Solution

Implement a centralized configuration system, similar to the options system in pandas or matplotlib. This would provide a single, global point of control for the default parameters of all analysis algorithms.

The proposed system would work as follows:

1. A central configuration object would be available, e.g., pkynetics.options.
2. Users could inspect and modify these global defaults in their scripts.

import pkynetics

# View all peak detection options
print(pkynetics.options.peak_detection)

# Globally change the default prominence for all subsequent analyses
pkynetics.options.peak_detection.prominence = 0.05

3. All functions in the library would be refactored to use these global defaults when a parameter is not explicitly provided by the user.

Alternative Solutions

• Status Quo: Continue with hardcoded defaults. (Rejected: Not scalable or reproducible).
• Configuration Files (YAML/JSON): Allow loading configurations from files. This could be an extension of the proposed system but adds complexity for a first implementation. The programmatic API should be prioritized.

Additional Context

This is a significant architectural improvement that will enhance the library's usability, maintainability, and scientific rigor by promoting reproducible analyses. Although it requires a considerable refactoring effort, the long-term benefits are substantial.

Implementation Details

1. Create a pkynetics/config.py module to house the configuration object and its API (set_option, get_option, etc.). A nested class or dictionary structure could be used.
2. Systematically refactor each module (e.g., dsc, dilatometry, model_fitting_methods) to use the new configuration system. The pattern would be:

   # Before:
   # def my_function(param: int = 21): ...
   
   # After:
   from pkynetics import options
   def my_function(param: Optional[int] = None):
       if param is None:
           param = options.some_module.param_default
       # ...

3. This work can be done incrementally, module by module, to make it manageable.

Guidelines

• I agree to follow this project's Contributing Guidelines
• I have searched for similar feature requests
• I understand that this is a request and implementation is not guaranteed