Skip to content

Implementation details

Implementation Details#

Some implementation detail on the state related classes is described in state-classes.pdf (source file: state-classes.odg).

Per State Information#

Use PerStateInformation<Entry> to store an Entry object for each state. We already discussed how to handle some special cases but not all of them are implemented yet (we will do so once we need them).

Temporary states/samples: Use a temporary state registry to create the samples.

PerStateInformation<Entry> is already able to handle states from different registries. Once the temporary registry is destroyed, all information stored for its states is also destroyed.

Create states from scratch: If states should be created that do not correspond to the initial state or the successor of an existing states, the state registry has to be extended. Add a method that allows to add a buffer (i.e. a state_var_t *) and handle it just like states in the get_successor_state() method.

Store information for only a few states: If information is needed for only some of the created states, consider using two registries: one for all states that require additional information and one for all other states. If this is not possible, create a new class similar to PerStateInformation<Entry> that uses a hash_map<StateID, Entry> instead of a SegmentedVector<Entry> for storage. Support for multiple state registries should be handled like it is done in PerStateInformation<Entry>.

States without duplicate checks: Depending on the use case it can make sense to either work directly with buffers (i.e. state_var_t *) or create a new state registry class that omits the duplicate check.