Skip to content

CHANGELOG.md

Latest commit

 

History

History
233 lines (160 loc) · 20.6 KB

CHANGELOG.md

File metadata and controls

233 lines (160 loc) · 20.6 KB

Changelog

All notable changes to RegressionLab are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.1.2 - 2026-03-07

Added

  • Extensible equation presets (YAML-first): Added equations.user.yaml for local user presets merged with base equations.yaml via a central equation_registry loader in config.
  • Expression-mode presets: Equation schema now supports type: expression (with expression, param_names, optional defaults/bounds) in addition to type: python, while keeping legacy YAML compatibility.
  • Security tests for custom formulas: Added tests to verify AST-based safe evaluation rejects unsafe payloads.

Changed

  • Equation output formatting: When a parameter is negative, the formatted equation no longer shows "+-" (e.g. y=5.2*x+-3.5). It now displays a single minus sign (e.g. y=5.2*x-3.5).
  • Fitting modularization: Refactored generic_fit into smaller helper functions (validation/preparation, curve fit execution, statistics, formatting) to improve readability and testability.
  • Equation resolution workflow: get_fitting_function now resolves both expression-based and python-target presets from the unified equation registry.
  • UI fallback behavior: Tkinter and Streamlit now show a readable equation label fallback when translations are missing, avoiding blocked onboarding for new presets.
  • Entry points: Consolidated namespaced entry points under regressionlab.* and aligned project scripts.

1.1.1 - 2026-03-05

Changed

  • Code cleanup (Maintainability): Removed dead code and unnecessary exports across the codebase. Cleaned up unused imports, eliminated redundant helper functions, and removed obsolete code paths that were no longer referenced.

1.1.0 - 2026-03-01

Changed

  • Equation lookups (Performance): get_equation_format_for_function() and get_equation_param_names_for_function() now use O(1) reverse-lookup dicts (_FUNCTION_TO_EQUATION, _FORMAT_TO_FORMULA) instead of scanning the entire EQUATIONS dict on every call. The formula lookup in generic_fit also uses the same O(1) dict. Affects every fit_* call.
  • Regex compilation (Performance): The ~40 regex patterns in MATH_FUNCTION_REPLACEMENTS are now pre-compiled at module load (MATH_FUNCTION_REPLACEMENTS_COMPILED). custom_function_evaluator._prepare_formula() uses the compiled patterns, avoiding recompilation on every custom formula evaluation.
  • Residuals reuse (Performance/Readability): In generic_fit, (y - y_fitted) ** 2 is now computed once as residuals_sq and reused for both R² and RMSE calculations.
  • Formatting cleanup (Readability): In fitting_utils.py, nested .format() calls for parameter text replaced with direct f-strings; merge_initial_guess now uses zip(computed, override) instead of range(len(computed)).
  • Transform dispatch (Performance/Maintainability): _apply_to_column in transforms.py refactored from a ~200-line if/elif chain to a dispatch dict (_TRANSFORM_DISPATCH) mapping transform IDs to small handler functions. Adding a new transform now requires only writing the handler and adding one dict entry.
  • Simplified _base.py re-exports (Maintainability): fitting/functions/_base.py now uses standard from module import name imports instead of manual tuple-assignment. Reduces update points when adding new estimators or utilities.
  • Walsh-Hadamard Transform (Performance): Replaced the O(n²) dense matrix approach (scipy.linalg.hadamard(n2)) with a vectorized in-place Fast Walsh-Hadamard Transform (FWHT) using only NumPy — O(n log n) time, O(n) memory. Removed scipy.linalg.hadamard dependency from transforms.
  • 3D plot fallback (Performance): The nearest-neighbor fallback in plot_utils.py (used when scipy is unavailable) now processes grid points in chunks, bounding peak memory to ~10M floats instead of allocating a full (grid_size² × n_points) distance matrix.

1.0.0 - 2026-02-14

Added

  • Data view transforms: New transforms in Watch Data / View Data mode: Laplace (discrete), inverse Laplace, Hilbert, inverse Hilbert, envelope (Hilbert), cepstrum (real), Hadamard (Walsh), inverse Hadamard. Useful for telecommunications, signal processing, and systems analysis. Transforms ordered by relevance (FFT/DCT first, then log/exp, standardize/normalize, Hilbert/Laplace, telecom).
  • Documentation (i18n): Sphinx i18n support for English and Spanish. ReadTheDocs can host both: create a separate project (e.g. regressionlab-es) with Language set to Spanish and link it as a translation of the main project. Spanish translations added for: index, introduction (complete), installation (partial), license (partial), api/index (partial), configuration (titles). Remaining .po files can be filled manually. See docs/README.md (Localization).
  • Update check (Tkinter): Weekly check for updates when opening the app. If a newer version is available and CHECK_UPDATES is enabled (default), a dialog asks whether to update. If confirmed, the app runs git pull; input/, output/, and .env are preserved. Configurable via .env: CHECK_UPDATES (enable/disable), CHECK_UPDATES_FORCE (force check every time, for testing), UPDATE_CHECK_URL (URL to fetch version from). See Configuration Guide.

Changed

  • Tkinter window positioning: All dialogs and secondary windows now open centered on screen. New frontend.window_utils module with place_window_centered(). Small dialogs keep their natural size; help and config dialogs use fixed dimensions (900×650 and 760×800 px).
  • Pair plots (View Data / Watch Data): Improved readability when there are many variables. Each cell now has a minimum size (1.4 in) so scatter plots stay legible. When there are more than 8 variables, Streamlit shows a multiselect to choose which variables to display (max 8); Tkinter opens a selection dialog. The selection is preserved when refreshing after transform/clean.
  • Imports and package structure: Centralized imports in all __init__.py files; __all__ now follows import order. Cross-package imports use the package root (e.g. from config import ..., from loaders import ...). Imports within the same package use full module paths (e.g. config.constants, loaders.loading_utils) to avoid circular references.
  • config: Added config.color_utils with lighten_hex, muted_from_hex, parse_hex_to_rgb; re-exported from config. Added get_env_from_schema to config exports. Streamlit theme now uses config.color_utils instead of duplicate implementations.
  • data_analysis: transforms and cleaning now use shared get_numeric_columns from data_analysis._utils (removed duplicate implementations).
  • main_program: Extracted _resolve_multiple_x_variables() helper to simplify repeated logic in fitting workflows.
  • Docstrings: Standardized to Google style; "Example:" → "Examples:" for consistency across modules.

0.9.3 - 2026-02-13

Added

  • Data view (Tkinter): New data analysis options in the "View data" window:
    • Transform: Apply FFT, FFT magnitude, inverse FFT, DCT (cosine), inverse DCT, log, log10, exp, sqrt, square, standardize (z-score), or normalize [0,1] to numeric columns. Dropdown to select transform; click Transform to apply.
    • Clean: Drop NaN rows, drop duplicates, fill NaN (mean/median/zero), or remove outliers (IQR or z-score). Dropdown to select cleaning; click Clean to apply.
    • Save: Save updated data button saves the current (possibly transformed/cleaned) data to CSV, TXT, or XLSX.
    • Help: Help button opens a dialog with collapsible sections (pair plots, transform, clean, save). Transform and Clean sections list each option with a detailed description. Window 900×650 px; Accept button fixed at bottom. Content in Spanish, English, and German.
  • Data view (Streamlit): Same transform and clean options in the "View Data" mode. Select from dropdowns, click Apply; download updated data as CSV, TXT, or XLSX. Help expander with pair plots, transform options (each detailed), clean options (each detailed), and save—same content as Tkinter, in Spanish, English, and German.
  • Data view (Tkinter): Pair plots window auto-updates when data is transformed or cleaned (if already open).
  • Module: New data_analysis package with transforms and cleaning submodules. Save logic in loaders.saving_utils; save dialog in frontend.ui_dialogs.save_data_dialog.

Changed

  • Data view window (Tkinter): Reduced height; transform and clean each in their own row; save button next to pair plots. Transform and Clean buttons use the same accent color as equation buttons (Equation.TButton).
  • Tkinter file loading: Replaced the two-step dialogs (file type + file name) with a single native file picker (filedialog.askopenfilename). Works on Windows, Linux, and macOS. Users can now browse and select any CSV, TXT, or XLSX file in one step.
  • Documentation: Updated Linux installation guide in docs/installation.md to reflect current behavior of install.sh and setup.sh:
    • Documented that install.sh clones into regressionlab (lowercase) and optionally installs Git on supported Linux distributions (apt, dnf, yum, zypper, pacman).
    • Documented that setup.sh can optionally install Python 3.12 and Tkinter on Linux when missing.
    • Updated prerequisites with additional package managers (zypper, pacman) and note about automatic Git installation.
    • Corrected installation folder path (regressionlab) for launching the app after quick install.

0.9.2 - 2026-02-12

Fixed

  • Hover: Minor hover bug fixed.

Changed

  • General: Significant code optimizations across loaders, paths, and fitting modules.
  • 2D fit plots: The fitted curve is now drawn by evaluating the fitted function on a dense linspace (300 points) over the full x range, instead of plotting only at the original data points. Produces smoother curves when fit_info is available.
  • 3D fit plots: The fitted surface is now drawn by evaluating the fitted function on a regular grid (50×50) of linspace points over the full x_0 and x_1 range, instead of interpolating the fitted values. Produces smoother surfaces when fit_info is available.
  • Loaders: Reader registry renamed from _READERS to FILE_TYPE_READERS and exported from loaders for reuse.
  • Paths: get_output_path now uses pathlib consistently instead of os.path.
  • Fitting: Simplified format_parameter logic; moved find_peaks import to module level in estimators.
  • Streamlit plot titles: The plot title now shows only the base name (e.g. "fit") without the numeric suffix (_1, _2). File paths still use the counter internally to avoid overwriting when making multiple fits in the same session.

Added

  • Configuration: New PLOT_SHOW_GRID environment variable (default false) to enable or disable a background grid on regression plots. Configurable via .env or the Tkinter configuration dialog. Applies to 2D fits, residual plots, and 3D plots.
  • Streamlit: Checkbox before each fit to toggle plot title visibility (show/hide). Default comes from PLOT_SHOW_TITLE environment variable; can be overridden per fit without changing the env.
  • Multidimensional fitting: Support for multidimensional regression. With 2 independent variables, an interactive 3D plot is shown; with more than 2 independent variables, a residuals plot is displayed.
  • Predictions window: New "Prediction" button in the result window opens a dialog to evaluate the fitted function at user-specified inputs. Supports one or more independent variables; displays the predicted value and propagates parameter uncertainty when available. Works for single/multiple datasets, multiple equations, and "all equations" mode.

0.9.1 - 2026-02-11

Fixed

  • Streamlit Cloud: Hotfix to repair Streamlit Cloud deployment that was not working due to a tkinter dependency that could not be resolved.

0.9.0 - 2026-02-07

Fixed

  • Binaries: Binaries corrected.

Changed

  • Python: Only Python 3.12+ is now supported.
  • Estimators: Estimator improvements.
  • Imports and maintainability: Improved imports and maintainability. New functions can now be defined more easily.
  • run.bat / run.sh: When using run.bat and run.sh, the command console is no longer shown.
  • User interface: Completely modernized UI across the application (layouts, styles, and controls updated).
  • User experience: Improved UX with better controls, configuration options, and visualizations for fitting and analysis.
  • Streamlit app: Refined Streamlit UI and behavior; current design is considered stable with no further near-term changes planned.
  • Documentation and docstrings: Many docstrings updated and expanded; user and API documentation completed and brought in line with the current behavior.
  • Environment variables: Improved loading, validation, and defaults when reading environment variables.
  • Dialogs: Dialog layout, wording, and behavior improved for a more consistent user experience.
  • Misc: Minor changes and optimizations.

0.8.3 - 2026-02-06

Fixed

  • Languages: Minor language fixes in translations and UI messages.

Added

  • .env validations: New validations for .env values to ensure correct configuration.
  • RMSE: Added RMSE (Root Mean Square Error) to fit results and statistics.

Changed

  • Python: Project requirement updated from Python 3.10+ to Python 3.12 only. All config (pyproject.toml, setup scripts, docs, installers) now require Python 3.12 or higher.
  • Tests: Tests updated to reflect recent changes and new validations.
  • Documentation: Docs submodules updated.
  • Imports: Improved relative imports across the project.
  • Fitting functions: Reference to existing functions moved to a .yaml file for better maintainability.
  • Environment variables: When invalid values are set in the environment variables, the system now detects them and applies the default values at the conflicting points.
  • Style and architecture: Minor style and architecture changes.

0.8.2 - 2026-02-04

Fixed

  • Closing with X: Corrected behavior when closing dialogs or the main window with the window close button (X). All Tkinter dialogs (equation type, file type, file name, variables, parameter configuration, custom formula, number of fits, config, data preview) now use WM_DELETE_WINDOW so that closing with X is treated as cancel and returns the appropriate signal. Main window closing with X shows the exit confirmation dialog (same as the Exit button); in the exit confirmation dialog, closing with X cancels exit.
  • Custom function unicode symbols: Fixed bugs where unicode symbols were not being read correctly in custom functions, ensuring proper parsing and evaluation of mathematical expressions with special characters.

Added

  • Parameter estimates and bounds: In Tkinter, you can configure initial parameter values and bounds before fitting via the "Configure initial parameters" dialog. Dialog columns: initial value, range start, range end per parameter; empty fields use the automatic estimator. All built-in fitting functions support initial_guess_override and bounds_override (with merge_initial_guess and merge_bounds utilities).
  • Extra information in results: Fit results can include additional statistical and fit information (e.g. χ², R², confidence intervals) in the result text and display.
  • Data visualization modes: Table view and pair plots (scatter matrix) for datasets. In Streamlit: expander "Data visualization" with dataframe and optional "Visualize variable pairs"; in Tkinter: data preview with optional pair-plot button that opens a scaled image window.
  • Configuration: Configuration dialog (Tkinter) with grouped sections: language, UI appearance, plots, fonts, paths, links, logging. All ENV_SCHEMA options are editable from the application and persisted to .env; closing the dialog with X cancels without saving.
  • Termux / mobile installer: First trial installer for Android via Termux (install_termux.sh). Installs Git and Python if needed; uses TUR repo and pkg for numpy, scipy, pandas, matplotlib and python-tkinter to avoid long pip builds. Clones the repo into ~/python_materials/RegressionLab, creates a venv (with --system-site-packages when using pkg packages), installs Streamlit and remaining dependencies, configures .env from .env_mobile.example, prompts for input/output paths, and creates a run script that starts the Streamlit app in headless mode (http://localhost:8501). Optional shortcut in ~/storage/downloads/run_regressionlab.sh when storage is set up. Recommended Termux from F-Droid.
  • TXT data files: Support for loading data from .txt files (tab or comma-separated; delimiter auto-detected). File type selector in Tkinter and Streamlit includes CSV, XLSX and TXT.
  • Plot output format: The output format for saved plots can be chosen in the configuration dialog (Tkinter) or via .env: FILE_PLOT_FORMAT with options png, jpg or pdf. Plots are saved in the selected format in the output directory.
  • Example datasets: Added more example datasets to the input/ directory for testing and demonstration purposes.

Changed

  • Config: src/config.py refactored into package src/config/ with modules constants.py, env.py, paths.py, theme.py. Public API preserved via config/__init__.py re-exports.
  • Fitting functions: src/fitting/fitting_functions.py split into src/fitting/fitting_functions/ and src/fitting/functions/. Model functions grouped by family in functions/: polynomials.py, trigonometric.py, inverse.py, special.py, plus _base.py for shared types.
  • Frontend (Tkinter): src/frontend/ui_dialogs.py refactored into package src/frontend/ui_dialogs/ with modules: config_dialog.py, data_selection.py, equation.py, help.py, result.py, tooltip.py.
  • Streamlit app: Application logic moved into src/streamlit_app/sections/ with modules: data.py, fitting.py, help_section.py, modes.py, results.py, sidebar.py. app.py delegates to these sections.
  • Imports: main_program.py, streamlit_app/app.py and utils/validators.py updated to use the new package layout.
  • Excel input: Support for legacy XLS (Excel 97-2003) has been removed; only XLSX is supported for Excel files.
  • Code refactoring: Continued refactoring of code and structures to improve maintainability, organization, and code quality.

0.8.1 - 2026-02-02

Fixed

  • Custom function fitting: Corrected failure when using custom functions due to parameter initialization. When *params was not provided, the code now initializes parameters with ones to handle the indeterminate number of parameters.
  • Tests: One test was fixed to align with the above behavior.

Added

  • German (Deutsch): Full locale support for German (de) added (translations in src/locales/de.json).
  • Statistical values: Improved syntax for including statistical values in results; new statistical values were added.
  • Donation / channel link: Optional donations button in the Help window (Tkinter and Streamlit). Configure via DONATIONS_URL in .env; if set, a button linking to that URL is shown. Documented in .env.example.

Changed

  • Documentation: Version references and docs updated for 0.8.1 (README, docs, Sphinx).
  • Configuration: DONATIONS_URL support added in config.py; optional in .env.example.

0.8.0

Initial 0.8.x release. See repository history and documentation for features and changes prior to 0.8.1.