DecisionModule does not accumulate state

[willu47]

We are dealing with system state in an inconsistent manner, which means that while pre-specified planning decisions are included in every year of the state file, outputs from previous years of a decision module are not rewritten to a state file.

Solutions:

TLDR; either write separately and read in a bundle, or read separately and write in a bundle;

1. only write pre-specified planning in base timestep state file and accumulate state through successive years for each decision iteration/timestep path followed, rewriting only the current decisions in each state file, reading them all in through the data_handle.get_state method
2. the statefile in each timestep holds the complete history, so only the most recent statefile is needed to see a history of all the interventions that took place in a particular timestep/iteration path.

[willu47]

In chat with Tom:

• data_handle.get_state should read state from three sources - initial conditions, pre-specified planning and decision_module decisions
• decision module needs to read the state in the previous year (consisting of only available interventions and accumulated from the base year) and write state for the current decision timestep

[willu47]

Methods to tidy up include:

data_layer/*/*_data_store.py

• read_strategy_interventions
• read_initial_conditions

data_layer/store.py

• read_strategies
• read_all_initial_conditions
• read_strategy_interventions

decisions/decision.py

1. Strip out all concept of multiple strategies and pre-specified planning - both initial conditions and pre-specified planning can be handled using the existing concept of state. They should be read in and filtered by the get_state method for the current timestep.
   • this may require interventions to drop lifetime attribute, and the state to be pre-computed so that an intervention is present in the state for each timestep during its lifetime
2. Decision module creates a register of available interventions. These should contain the set of interventions minus the subset of interventions installed in pre-specified and initial condition states.
3. Decision module requires access to previous timestep/iteration state. (results_handle should provide read only access to state from previous timestep/iterations #334)
4. DecisionManager get_and_save_decisions writes state for the current timestep which excludes pre-specified and initial_conditions