Skip to content

Changelog

All notable changes to GemsPy are documented here.

[0.1.2] - 2026-06-11

Added

  • Properties in Models and Components - models in library YAML files may declare an optional properties list (entries with an id). Components in system YAML files carry optional properties.
  • Taxonomy - new gems.model.taxonomy module to represent Taxonomy, i.e. a classification of GEMS models. Optional taxonomy-category on models in library YAML files, exposed as ModelSchema.taxonomy_category.
  • Math operators abs and round in the GEMS expression language.
  • Can be applied to parameters and literals in constraints, bounds, and objective contributions (degree-0 context).
  • Can be applied to any expression in extra-outputs (post-solve evaluation), including decision variables.
  • Use .abs() and .round() methods on expression objects, or abs(expr) and round(expr) in parsed expression strings.
  • dual(constraint_name) and reduced_cost(variable_name) operators in the expression language, usable in extra-outputs. Both are validated at model-build time; using them inside constraints or objective contributions raises a ValueError.
  • Xpress (≥ 9.8) and Gurobi (≥ 10.0) solver support alongside HiGHS. Reduced costs use each solver's native API (getLpSol / getAttr("RC") / col_dual) since linopy has no unified reduced-cost interface.
  • New solvers optional dependency group (uv sync --group solvers) for running solver-specific tests with a licence; CI now installs it.

Fixed

  • Pre-commit black hook no longer passes redundant --config pyproject.toml.
  • Pre-commit isort hook simplified to --profile black (removed --filter-files which conflicted with pre-commit's own file filtering).

Changed

  • Modernized README design: new layout, GEMS favicon, quick-link navigation, and uv install instructions.

[0.1.1] - 2026-05-29

Scenario-scope playlist (replaces nb-scenarios)

The scenario-scope section of optim-config.yml now supports a full playlist mechanism. The old nb-scenarios integer key is removed and raises a validation error if still present.

Scenario indices are 0-based throughout, consistent with the modeler-scenariobuilder.dat convention.

Inline form — specify scenarios with integers, string-integers, and inclusive "a-b" range strings:

scenario-scope:
  include:
    - "0-49"
    - 74
    - "89-99"
  exclude:
    - 9
    - 14

Playlist-file form — point to a flat JSON array of 0-based integers, useful for machine-generated playlists:

scenario-scope:
  playlist-file: mc_playlist.json

Other changes:

  • exclude is now compatible with both include and playlist-file. Use it to subtract a few scenarios at run time without modifying the playlist file.
  • validate_optim_config() now accepts an optional scenario_builder argument and cross-checks all playlist indices against every scenario group, raising a ValueError for out-of-bounds indices.
  • The playlist is resolved and cached exactly once at load_optim_config() time; I/O and format errors surface immediately as ValueError.
  • Boolean values (true/false) are explicitly rejected in both inline lists and JSON playlist files.

[0.1.0] - 2026-04-30

Study folder structure

Studies are now loaded from a conventional directory layout via load_study(study_dir):

<study>/
  input/
    system.yml                          # component topology and connections
    model-libraries/*.yml               # model library files
    data-series/                        # time-series parameter files (TSV)
    data-series/modeler-scenariobuilder.dat   # optional scenario mapping

The Study object (pairing a System with a DataBase) is the single input to SimulationSession.

Optimization configuration (optim-config.yml)

A new optim-config.yml file controls all aspects of a simulation run:

  • resolution.mode — four strategies: frontal, sequential-subproblems, parallel-subproblems, benders-decomposition
  • resolution.block_length / block_overlap — time-window size and overlap for sequential/parallel modes
  • time_scope — first_time_step / last_time_step
  • scenario_scope.nb_scenarios — number of Monte-Carlo scenarios to run (replaced by the playlist mechanism in a later release)
  • solver_options — solver name (default: HiGHS), log verbosity, and free-form solver parameters
  • models[].model_decomposition — per-model assignment of variables, constraints, and objective contributions to master, subproblems, or master-and-subproblems (used for Benders decomposition)
  • models[].out_of_bounds_processing — per-constraint handling of time indices that fall outside the horizon (cyclic or drop)

Scenario builder

modeler-scenariobuilder.dat maps Monte-Carlo scenario indices to data-series column indices on a per-scenario-group basis, allowing different components to draw from different stochastic draws.

Vectorized solver

The internal optimizer was migrated from a scalar OR-Tools pipeline to a fully vectorized linopy / xarray pipeline. All constraints for all components, time steps, and scenarios are now built in a single pass and solved in one call — significantly reducing build time for large horizons.

Removed

  • OR-Tools dependency.

[0.0.6] - 2025

Added

  • Math operators floor, ceil, min, and max in the GemsPy expression language, with max/min accepting a variadic argument list (argList grammar rule).
  • Support for complex variable bounds expressions: operator expansion now runs before bounds evaluation, enabling expressions such as min(p_max_cluster, min_gen_mod * unit_count * p_max_unit).
  • visitTimeShiftExpr and visitTimeIndexExpr in the expression visitor interface.
  • AGENTS.md and agent guidance documentation.
  • Status and quality badges in README.

Fixed

  • O(T²) list-copy performance bug in CopyVisitor.addition (now O(T)).
  • Pandas and Pydantic deprecation warnings resolved.
  • floor/ceil/max/min visitor implementations in LinearExpressionBuilder.
  • Degree of floor/ceil/max/min returns inf for non-constant operands.

[0.0.5] - 2025

Added

  • load_input_system: load and validate an input system into memory from a file path.

[0.0.4] - 2025

Removed

  • PyPSA → GEMS converter and related files (out of scope).

[0.0.3] - 2025

Changed

  • Version naming aligned; package prepared for PyPI upload.
  • README revised to consistently use GEMS / GemsPy naming.

[0.0.2] - 2025

Changed

  • Initial PyPI preparation: project metadata, classifiers, and packaging configuration.