Changelog#

All notable changes to CVXlab will be documented in this file.

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

[Unreleased]#

variable_domain field for data tables: replaces the old boolean integer field. Accepted values: integer (integer variables, $\mathbb{Z}$) or boolean (binary variables, ${0,1}$). Omitting the field keeps the default continuous domain. The deprecated integer: true syntax is still accepted with a warning and automatically migrated to variable_domain: integer.
Installation docs: new Install from Dev Branch section with git+https install instructions for users who want the latest unreleased features.
Possibility to run only a selection of scenarios from Model.run_model() method, by specifying the scenario_idx attribute as integer or list of integers corresponding to the index of scenarios in the Model.scenarios property.
Per-problem solver and solver-settings routing for multi-problem runs (backend: core.py, problem.py, database.py).

Reorganized backend solve flow; moved database comparison/cleanup into dedicated helpers (backend/database.py, backend/core.py).
Split nonlinear tutorial assets into separate model_nonlinear and model_decomposition sets under docs/source/tutorials/production_planning_nonlinear.
Bumped minimum cvxpy version to >=1.9.1 (see pyproject.toml).
Adjusted numerical default: Defaults.NumericalSettings.SPARSE_MATRIX_ZEROS_THRESHOLD changed from 0.3 to 0.7 (defaults.py).

Backward-compatibility migration for integer field: the migration block that converts integer: true → variable_domain: integer.
README images now use absolute raw GitHub URLs so they render correctly on PyPI.
Corrected GitHub organization URL (cvxgrp → cvxlab) throughout installation docs.
Replaced unsupported tab-set/tab-item directives (sphinx-design) with plain RST in the citation section of the resources page.
Integration tests and CI: integration tests now run across tutorial models and were updated to cover solver routing and tutorial assets (tests/integration/test_integration.py).

Frontend CLI (cvxlab.frontend.run()): interactive menu-driven interface for model setup and execution. Modules: interface, session, actions, display.
User-defined constants and operators: template modules (user_defined_constants.py, user_defined_operators.py) for custom symbolic extensions; auto-imported when present in the model directory (issue #109).
CSV input data support: models can now use CSV files as input data sources (issue #103).
Model.update_sets_tables(): update set definitions in an existing SQLite database without full regeneration.
skip_tables argument in Model.run_model() / Core.solve_integrated_problems(): selectively exclude tables from convergence checks.

Python ≥ 3.11 required: bumped requires-python; removed from __future__ import annotations across the codebase.
Integrated-solving convergence refactored: reworked algorithm in Core.solve_integrated_problems().
Centralized Defaults.LiteralTypes: shared Literal type aliases consolidated, replacing scattered definitions.
Identity matrix constant: now accepts a single set as dimensional argument (issue #101).
description columns in data structures are no longer processed, allowing free text (issue #105).
Removed CRITICAL log level from Logger.
Pinned pandas==2.3.3; removed deprecated errors='ignore' arguments (pandas 3.0 compatibility).

util.normalize_dataframe(): improved NaN handling and blank-fill logic.
SQLManager.dataframe_to_table(): string-type conversion and batch flag reset.
Database.load_data_input_files_to_database(): input directory path reference and NaN replacement before SQLite export.
Database.generate_blank_data_input_files(): corrected excel_dir_path argument.
Model.run_model(): selected solver now correctly logged.
util.pivot_dataframe_to_data_structure(): handle missing primary key column in Excel setup files.
Various error-catching improvements for symbolic expression validation and settings inconsistencies (issues #102, #104, #106, #107, #108).