v0.12.0

What's Changed

• Large update (ease-of-use, parametric models, MEC elimination, static visualisation of models via dot, teaching module) by @sjunges in #484
• replace lxml with standard xml handling by @sjunges in #495
• some mild refactoring in the docs by @sjunges in #492
• Make pygame optional by @sjunges in #494

Full Changelog: v0.11.0...v0.12.0

v0.11.0

This is a large refactor of stormvogel. It provides a much more stable and predictable Model API.

Breaking Changes

Below is a comprehensive list of breaking changes that affect users upgrading to this version.

Module layout

The single stormvogel/model.py file has been split into a package stormvogel/model/ with submodules (model.py, state.py, action.py, choices.py, distribution.py, observation.py, reward_model.py, value.py, variable.py). All public names are re-exported from stormvogel.model, so import stormvogel.model continues to work.

bird builder

Old	New
rewards accepted (state, action) for action models and (state) for non-action models	rewards must always accept exactly one parameter: (state)
valuations callback returned dict[str, ...]	valuations callback must return dict[Variable, ...]

Model API

Attribute changes

Old	New
model.type	model.model_type
model.states: dict[int, State]	model.states: list[State]
model.choices: dict[int, Choices]	model.transitions: dict[State, Choices]
model.actions: set[Action] | None (attribute)	model.actions (property, yields from transitions)
model.exit_rates: dict[int, ValueType] | None	Removed — rates are encoded in transition values

Getters replaced by properties

Old	New
model.get_initial_state()	model.initial_state
model.get_variables() returned set[str]	model.variables returns set[Variable]
model.get_parameters()	model.parameters

model.get_labels() (which returned set[str]) has been removed. The closest equivalent is model.state_labels, but note the type changed: it is a dict[str, set[State]] mapping each label to its states. Use set(model.state_labels.keys()) to get the old behavior.

Renames

Old	New
model.set_choice(state, choices)	model.set_choices(state, choices)
model.add_choice(state, choices)	model.add_choices(state, choices)
model.parameter_valuation(values)	model.get_instantiated_model(values)
choice_from_shorthand(shorthand, model)	choices_from_shorthand(shorthand) (no model parameter)

Signature changes

Old	New
for id, state in model:	for state in model:
model[state_id] (by int id)	model[index] (by list index)
model.get_state_by_id(int)	model.get_state_by_id(UUID) — for integer index access use model.states[i]
model.get_states_with_label() returned list[State]	Returns set[State]
model.get_sub_model(states: list[State])	model.get_sub_model(states: Iterable[State])
model.iterate_transitions() returned list	Returns Iterator[tuple[ValueType, State]]
model.unassigned_variables() returned list	Returns Iterator[tuple[State, Variable]]
model.get_successor_states(state) returned set[int]	Returns set[State]
model.remove_state(state, normalize, reassign_ids)	model.remove_state(state, normalize, suppress_warning) — reassign_ids removed
model.make_observations_deterministic(reassign_ids)	model.make_observations_deterministic() — reassign_ids removed
model.observation(observation_id: int, valuations)	model.observation(alias: str) — create-or-get by alias
model.new_observation(valuations, observation_id)	model.new_observation(alias: str)
model.new_state(..., name=) accepted a name argument	The name parameter has been removed

New methods

Method	Notes
model.get_state_index(state) -> int	O(1) amortized index of a state in model.states
model.get_distribution(state) -> Distribution	Replaces model.get_branches(state)
model.get_observation(alias: str) -> Observation	Strict lookup; raises KeyError if not found
model.new_reward_model(name: str) -> RewardModel	Factory for reward models
model.get_default_rewards() -> RewardModel	Returns the first reward model
model.get_rewards(name: str) -> RewardModel	Looks up a reward model by name

ModelType.HMM has been added alongside POMDP. Use new_hmm() to create an HMM model.

Removals

The following methods have been removed without a direct replacement:

• model.get_state_by_name() — use labels to identify states.
• model.get_action() and model.get_action_with_label() — use model.action("label_string") instead.
• model.get_choices(state) and model.get_choice(state) — use state.choices instead.
• model.get_states() — use model.states directly.
• model.get_ordered_labels() — build from model.state_labels if needed.
• model.get_rate() and model.set_rate() — rates are now encoded in transition values.
• model.get_choice_id() and model.get_state_action_id().
• model.get_state_action_pair().
• model.remove_transitions_between_states().
• model.all_non_init_states_incoming_transition().
• model.reassign_ids().
• model.has_only_number_values().
• Model.__eq__ — model comparison now uses identity (is), not structural equality.

State API

Identity and hashing

Old	New
state.id: int	state.state_id: UUID
@dataclass() with custom __eq__ comparing integer id, labels, valuations	@dataclass(eq=False) — identity-based equality
Unhashable (could not be used as dict key)	Hashable by identity (used as dict key throughout)

Attributes and properties

Old	New
state.labels was a list[str] attribute	state.labels is a property returning Iterable[str] (backed by model.state_labels)
state.get_choices() / state.get_choice()	state.choices (property)
state.get_observation() / state.set_observation()	state.observation (property with setter)
state.nr_choices() (method)	state.nr_choices (property)
state.add_valuation(variable: str, value)	state.add_valuation(variable: Variable, value)
state.get_valuation(variable: str)	state.get_valuation(variable: Variable)
(absent)	state.friendly_name -> str | None (property)
(absent)	state.set_friendly_name(name)

Renames

Old	New
state.set_choice(choices)	state.set_choices(choices)
state.add_choice(choices)	state.add_choices(choices)

Choices API

The Branches class has been removed. Each action in a Choices now maps directly to a Distribution[ValueType, State] object.

Old	New
choices.choice: dict[Action, Branches] (attribute)	Use choices[action] or for action, branch in choices:
choices.actions() (method)	choices.actions (property)
Branches.branch: list[tuple[Value, State]]	Distribution[Value, State] (see Distribution API below)
Branches.get_successors() returning set[int]	Distribution.support (property) returning set[State]
Branches.sum_probabilities()	Removed
Choices.sum_probabilities(action)	Removed
Choices.__eq__, Branches.__eq__ (structural)	Removed — identity-based equality

Distribution API (new, replaces Branches)

Branches has been replaced by the generic Distribution[ValueType, SupportType] class. A Choices object now maps each Action directly to a Distribution[ValueType, State].

Feature	API
Construction from list	Distribution([(prob, state), ...])
Construction from dict	Distribution({state: prob, ...})
Iteration	for value, state in distribution:
Support set (replaces get_successors())	distribution.support → set[State]
Probability list	distribution.probabilities → list[ValueType]
Stochasticity check	distribution.is_stochastic(epsilon)

RewardModel API

Old	New
RewardModel.rewards: dict[tuple[int, Action], ValueType]	RewardModel.rewards: dict[State, ValueType]
get_state_reward(state) raised on action models	get_state_reward(state) works on all models
set_state_action_reward(state, action, value)	Removed. Use set_state_reward(state, value).
get_state_action_reward(state, action)	Removed. Use get_state_reward(state).
set_from_rewards_vector(vector)	set_from_rewards_vector(vector, state_action=False). When state_action=True, the vector is interpreted as one entry per (state, action) pair.
RewardModel.__eq__ (structural)	Removed — identity-based equality

Observation API

Old	New
Observation.observation: int	Observation.alias: str with a UUID observation_id

Interval fields

Old	New
Interval.bottom	Interval.lower
Interval.top	Interval.upper

Variable type

State valuations previously used plain str keys; they now use Variable objects (a frozen dataclass with a label: str field, e.g. Variable("x")). All valuation-related APIs in State and Model are affected — see the respective sections above.

v0.10.2

What's Changed

• Updated links to new repository by @volkm in #467
• Improved performance of model mapping to stormpy by @volkm in #468

Full Changelog: v0.10.1...v0.10.2

v0.10.1

What's Changed

• stochastic observations and generic types by @lukovdm in #459

Full Changelog: v0.10.0...v0.10.1

v0.10.0

Breaking changes:

• Actions now use a single string label instead of a frozenset of labels (this affects the PGC builder!)
• Class names Branch and Choice are renamed to Branches and Choices

What's Changed

• Review fixes ivo by @PimLeerkes in #437
• Review fixes ivo by @PimLeerkes in #439
• Review fixes ivo by @PimLeerkes in #441
• Improved property builder by @YouGuessedMyName in #443
• change values in pomdp notebook to a dict by @YouGuessedMyName in #445
• Update docs by @YouGuessedMyName in #446
• update docs by @sjunges in #451
• Add LTS Layout that hides probabilities by @linusheck in #456
• Add docs/ folder to autoformat by @linusheck in #457
• Actions are strings by @linusheck in #458

Full Changelog: v0.9.10...v0.10.0

v0.9.10

What's Changed

Full Changelog: v0.9.9...v0.9.10

v0.9.9

What's Changed

• Review fixes ivo by @PimLeerkes in #421
• review fix bird.py by @PimLeerkes in #422

Full Changelog: v0.9.8...v0.9.9

v0.9.8

What's Changed

• Faster import by @linusheck in #418

Full Changelog: v0.9.7...v0.9.8

v0.9.7

What's Changed

• Add model analysis back to docs by @linusheck in #410
• Vis issues by @YouGuessedMyName in #416
• Update 4-Building-pomdps.py by @sjunges in #402

Full Changelog: v0.9.6...v0.9.7

v0.9.6

What's Changed

• fix duplication bug when using layout editor by @YouGuessedMyName in #396
• Extend show api and fix bugs related to the layout editor and saving/loading by @YouGuessedMyName in #408

Full Changelog: v0.9.5...v0.9.6