Skip to content

Changelog

Every release is listed here with its changes, grouped by type. Releases are prepared automatically from the commit messages, so the entries below are always in sync with the code.

How releases are made

Commits to main follow Conventional Commits. The type decides the next version number:

Commit type Example Effect while the version is 0.x
fix: fix: keep elites when the population shrinks patch (0.1.0 to 0.1.1)
feat: feat: add Boltzmann selection minor (0.1.0 to 0.2.0)
feat!: or a BREAKING CHANGE: footer feat!: rename Problem.fitness minor (0.1.0 to 0.2.0)
docs:, test:, chore:, ci:, refactor: docs: explain elitism no release by themselves

release-please keeps an open pull request that bumps the version (in pyproject.toml and carmelolg.galib.__version__) and adds the new entry to CHANGELOG.md. Merging that pull request creates the git tag (v0.2.0) and the GitHub release.

Releases

Changelog

0.3.1 (2026-11-18)

Documentation

  • add logo to README (3e1350d)
  • put logo and name side by side in README (122cdb6)

0.3.0 (2026-11-18)

Features

  • engine: parallel fitness evaluation and thinned history (db920e0)

Documentation

  • link PyPI package, drop not-on-PyPI note, ignore generated files (da4abe1)

0.2.2 (2026-11-17)

Bug Fixes

  • docs: make the mobile drawer navigable and touch friendly (#4) (ce7e092)

0.2.1 (2026-11-17)

Documentation

  • document results, metrics and plotting with examples (535873a)

0.2.0 (2026-11-17)

Features

  • initial release of GALib (4095171)

Documentation

  • drop the hand-written intro from the generated changelog (1ae3ef2)

0.1.0 (2026-11-17)

Features

  • initial release of GALib (4095171)

What 0.1.0 contains

The first version ships these features.

  • Problem base class: random_genome plus fitness or fitness_batch (vectorised evaluation), optional random_gene, crossover, mutate and maximize.
  • GeneticAlgorithm engine with tournament selection, crossover rate, per-gene mutation rate (default 1 / genome length), elitism, seeded and independent runs, optional executor for parallel fitness, and overridable building blocks (initial_population, evaluate, select, crossover, mutate, reproduce, replace, measure_diversity, collect_stats); initialize and step to drive the loop manually.
  • Selection operators: Tournament, Roulette, Rank.
  • Crossover operators: OnePoint, TwoPoint, Uniform; for permutations OrderCrossover, PartiallyMappedCrossover; for real values BlendCrossover.
  • Mutation operators: RandomReset; for permutations SwapMutation, InversionMutation; for real values GaussianMutation. Mutated positions are drawn with geometric jumps (cost proportional to the number of mutations, not to the genome length).
  • Optional numpy support: numpy array genomes in every built-in operator and metric, vectorised fast paths.
  • Diversity measures positional_diversity (default) and gene_spread, selectable with diversity=.
  • Stop conditions TargetFitness, Stagnation, TimeLimit; callbacks Callback, PrintProgress (flush, end, in_place, overridable format_line); plot_history (optional matplotlib).
  • Examples (phrase, knapsack, TSP, Rastrigin), numpy benchmark, documentation site sources (MkDocs Material), CI workflows.
  • Apache-2.0 license with NOTICE file.