Evolution of the fingerPro Package

Comparative analysis of the fingerPro package versions back to the original standalone C program. Each phase documents the state, features, and development metrics of the package, presented in reverse chronological order.

fingerPro 2.1

API Refinement and CRAN Preparation

This version focuses on standardizing the package for CRAN, refining the public API, and improving data ingestion workflows.

Key Features

• Parameter Standardization: The seed parameter is renamed to rng_init across the entire package (unmix, CR, CTS_explore, CTS_select, CI, etc.) to clearly indicate its role in initializing the random generator.
• Semantic Renaming:
  • CTS_seeds() is now CTS_explore() to better describe the exploration of tracer combinations.
  • CTS_error() is now CTS_select(), featuring an improved interface that returns a filtered dataset.
• New Utility Functions:
  • read_database(): A centralized entry point that handles file reading, validation, mixture selection, and zero-value detection.
  • validate_results(): Diagnostic function for post-unmixing analysis.
• Function Visibility: Internal validation functions are now hidden from the public API using @keywords internal.

fingerPro 2.0

Refinement and Variability Propagation

A major development cycle aimed at polishing the package for CRAN. While the initial objective was to include the Conservative Balance (CB) method, the project evolved into a deep revision of almost all existing functions and documentation.

Key Features

• Advanced Unmixing API: Supports variability (SD vs SEM), Linear Variability Propagation (lvp), and Latin Hypercube resolution.
• C++ Backend Expansion: Core C++ code includes unmix_c_lvp and variability correction helpers (correct_sem and correct_sd).
• Technical Metrics:
  • 900 lines of code added or modified.
  • 500 lines of new documentation and help files.
• Workflow: Transitioned to a Pull-Request (PR) based development model to facilitate collaborative updates.

fingerPro 1.4

**Restructuring and Standardization

This phase modernized the codebase by gathering dispersed functions and adopting consistent naming conventions.

Key Features

• Naming Convention: Transitioned all files and functions to snake_case (e.g., box_plot.R, DFA_test.R).
• Method Generalization: CR and CTS logic rewritten to support any number of sources (previously limited to 3 or 4).
• New Methods: Introduced Conservative Balance (CB_method) for isotopic tracers (initial implementation).
• Technical Metrics:
  • 2,400 lines of code added or modified.
  • 560 lines of new documentation.

fingerPro 1.3

Major Expansion: Selection and Logic

A significant growth phase that introduced the core tracer selection framework.

Key Features

• Codebase Size: Total of 3,200 lines of R code (800 of which are documentation) and 1,620 lines of C code.
• Tracer Selection Suite: Introduced Consensus Ranking (CR), Consistency Error (CTS), and the Conservativeness Index (CI).
• Analytic Solving: Added C++ support for least_squares_c and “triangles” methods.
• Pre-calculated Data: Added support for pre-averaged data (mean/SD/n) via the Means parameter.

fingerPro 1.2 / 1.1

Maintenance and Metadata * 1.2: Updated README.md to include CRAN/Zenodo badges and formal DOI citations. * 1.1: Maintenance release; identical to the 1.0 code base.

fingerPro 1.0

The Full R Package Transition

This version marked the transformation from a single-function bridge into a comprehensive statistical toolset. * Algorithm: New optimization using Latin Hypercube sampling with iterative variability corrections. * Statistical Testing: Built-in functions for Kruskal-Wallis (KW) and Discriminant Function Analysis (DFA). * Graphic Suite: Implemented a wide range of plots using ggplot2, including PCA biplots, LDA plots, and contribution density plots.

fingerPro 0.9

The Rcpp Bridge

The first iteration as an R package, originally named fingerprinting. It provided a single Rcpp wrapper around the original C algorithm, implementing a Goodness of Fit (GOF).

fingerPro-C 0.8

Standalone C Program

The original implementation of the unmixing logic. * Interface: Command-line execution using input.txt and solution.txt. * Dependencies: GNU Scientific Library (GSL). * Algorithm: Random weight generation (sorted uniforms) with quadratic GOF optimization.