model.roottol¶

The time roadrunner reports the rows of an event of the model at.

libroadrunner reports an event of the model in the steps of the integrator (variable_step_size) as two rows: the state at the root of its trigger, before the event, at root - roottol, and the state after the event at root + roottol; an event which roadrunner finds at the end of a step of CVODE rather than as a root has its first row at the end of the step less the machine epsilon (CVODEIntegrator::integrate of libroadrunner 2.10). A first step of CVODE after an event which ends within roottol of it is not reported, a second one is, after the row at root + roottol, so the times are not sorted. roottol is a static const of the function, which the first call of the process fixes to 100 * 32 * eps * (|t0| + |h|) of that call, t0 and h the start and the step of the first integration of the process, of any model: the steps of the integrator over 0..10000, whose first call integrates the whole interval, fix it to 7.1e-9, roadrunner's default simulate(0, 10000), 51 output times 200 apart, to 1.42e-10. After the first of them the rows of every event are 7.1e-9 before and after it, more than the resolution of the time of a simulation over 0..2.5, and a worker of a pool, whose first integration is its own, has other rows than the process which runs the scan.

root_tolerance runs a probe once per process, before the first model sbmlsim loads (RoadrunnerSBMLModel) and when a worker of a pool starts (sbmlsim.parallel), unless the main module, which a worker imports first, integrates when it is imported: a model with one event, whose first integration spans PIN_SPAN, so roottol is 6.2e-31 in every process in which sbmlsim integrates first, the two rows of an event have the time of the event and the first step of CVODE after it is reported. The probe then measures roottol from the rows of its event, also where an integration before fixed it, and at_event_times gives the two rows of every event the time of the event; the executor restores the rows of an event less than roottol before or after the end of a segment, whose row after it roadrunner folds into its last row, see sbmlsim.simulator.executor._events_at_the_end. In a process in which roadrunner integrated before sbmlsim loaded a model, the steps inside a segment lack the first step of CVODE after an event if it is shorter than that roottol, which roadrunner does not report, and the rows of an event which CVODE finds in that step, which roadrunner applies without rows; a worker of a pool has both, so the rows differ by them and the values on a grid by the interpolation across them. The probe warns once.

A roadrunner whose roottol is no constant of the process, which the probe cannot measure or which reports both rows of an event at its time degrades without an error and without a warning: the probe gives a roottol of 0 and the rows of an event stay at the times roadrunner reports them, which a time of a grid sees as a chain of jumps closer than the resolution of the time, see sbmlsim.result.timecourse.grid_weights. The behavior of libroadrunner 2.10 which sbmlsim depends on is pinned by the tests a new release of roadrunner has to pass: tests/model/test_roottol.py (the pin, the measurement and the warning), tests/simulator/test_event_rows.py (the rows of events after a first integration with roadrunner, in fresh processes), tests/simulator/test_executor.py (the events at the end of a segment and of a simulation, the folded rows, the rounding of the output times by roadrunner and the runs again past the end) and tests/fit/test_data_at_change.py (the data at an event).

RootTolerance(value, precision)

root_tolerance()

slack(earlier, later, tolerance)

at_event_times(time, tolerance)