Equivalent of the Python engine's RInterface. State is held
time-major (one environment per tick, each holding every channel); results
are patch-major ([npatches, nticks]). The three trimming rules are not
interchangeable and are reproduced exactly from model.py:82-165:
Usage
sim_results(state, par)
SIM_CHANNELS_TRIM_FIRST
SIM_CHANNELS_TRIM_LAST
SIM_CHANNELS_TRANSPOSE_ONLY
SIM_CHANNELS_PRECOMPUTED
SIM_CHANNELS_PASSTHROUGH
SIM_CHANNELS
SIM_CHANNELS_INTEGER
SIM_CHANNELS_DOUBLEArguments
- state
State environment from the run loop.
- par
Parameters from
sim_params().
Details
compartments, incidence, hazards,
N,Wdrop thet = 0seed row ([1:, :].T);event counts (births, deaths, reported) drop the last row (
[:-1, :].T), because they are written attickrather thantick + 1;doses and the precomputed beta/delta matrices are already
nticks-shaped and are only transposed.
Storage mode is per field: counts are integer, rates and hazards are
double. No dimnames – the Python return has none and no consumer
reads any.
