pyserq.Observe

Statistics and sample records for one observe declaration, or for one of the client's metrics. Obtain it from Report.observe(name), Report.observes or Report.client. Observe() is not a public constructor; attributes are read-only.

Attributes

Attribute	Type	Description
name	str	Observation name.
count	int	Number of recorded samples after warm-up.
mean	float	Sample mean.
ci	float	Batch-means 95% confidence-interval half-width; positive infinity below 40 samples.
cv2	float	Squared coefficient of variation.
p99	float	99th percentile of the recorded values.
samples	list[float]	Values in recording order.
times	list[float]	Simulation time of each sample.
sessions	list[int]	Session serial of each sample.
turns	list[int]	Turn number of each sample.

The four lists have matching positions. Each property access creates a new list; bind it once when processing many records. A CI is a half-width, not a pair of endpoints. When finite, its interval is mean ± ci. Statistics without a finite value remain NaN or infinity in Python.

Examples

import pyserq

source = """
fn main() {
  pool reqs { cap 2; }
  stage svc : fifo;
  workload {
    arrive batch(4);
  }
  server {
    set t0 = now;
    hold reqs (cost(reqs, 1)) { run svc (cost(svc, 2)); }
    observe response = now - t0;
  }
  gauge occupied = used(reqs);
}
"""
report = pyserq.run(pyserq.compile(source=source, horizon=10))
observation = report.observe("response")
assert observation is not None
records = list(zip(
    observation.times, observation.sessions,
    observation.turns, observation.samples,
))
assert len(records) == observation.count == 4

See also

Report, Gauge, observe statement.