Statman is a collection of metric collectors to embed within your python application. It includes a registry to easily access your metrics.
Statman => registry
Metric => set of classes that can perform metric collection
Stopwatch => a metric class responsible for tracking time delta
Gauge => a metric class responsible for providing a single value
Calculation => a metric class responsible for performing calculations
Rate => a specialized calculation metric which calculates x/y rate
Statman is availble from pypi.
It can be manually installed by:
pip install statman
or by adding the following to your requirements.txt:
statman=*
Statman offers a registery to make it easily to globally access metrics. Perhaps you will create and register a stopwatch in the depths of your codebase to measure the time to write to a database, and then want to access that result in some other part of your application.
register(name, metric)=> manually register a new metric
get(name)=> get a metric by name
count()=> returns a count of the registered metrics.
reset()=> clears all metrics from the registry.
stopwatch(name)=> returns a stopwatch instance. If there is a registered stopwatch with this name, return it. If there is no registered stopwatch with this name, create a new instance, register it, and return it.
Stopwatch is for timing operations within your system. Suppose that you are trying to track down where the system is slow. Put a stopwatch around certain critical areas, time those operations, and compare.
Stopwatch(name=None, autostart=False, initial_delta=None)=> create an instance of a stopwatch.- If
autostartset to true, the stopwatch will automatically start - If
initial_deltais set to a value, andreadof the stopwatch is incremented by this amount. This can be helpful if you adding timings together. nameis used for to string / reporting for identification of this metric. Defaults to blank- If
enable_historyis set to true, when a timing is collected (stopinvoked), an event is collected. This can be accessed by thehistoryproperty to examing statistics on this stopwatch
- If
start()=> starts the stopwatch, let the timing begin!
read(units, precision)=> reads the stopwatch to determine how much time has elapsed. Returns the time elapsed in seconds.- The elapsed time will be returned based upon the
units('m' minutes, 's' seconds, 'ms', milliseconds). Defaults to seconds. - If precision is provided,
read()will round to the number of decimals places based on precision. - Note:
readdoes NOT stop the stopwatch - if the stopwatch is runnning, it will continues to run.
- The elapsed time will be returned based upon the
time(units, precision)=> alias forread()
stop(units, precision)=> stops the stopwatch, and returns the time elapsed in seconds- See read for the role of
unitsandprecision
- See read for the role of
reset()=> restores the stopwatch back to init state and clears start and stop times
restart()=>resets the stopwatch, thenstarts it
history=> ifenable_historyset during stopwatch construction, thehistoryproperty returns an instance of a history object, which can be used for examing statistics
A gauge is an instantaneous measurement of a value. Suppose that you are interested in counting the number of messages that have been processed. A gauge can be used to count events and produce a value.
Gauge(name=None, value: float = 0)=> create an instance of a gauge- If
valueis provided, this will be used as the initial value of the gauge
- If
value()=> get / set the current value of the gauge
increment(amount: int = 1)=> adds to the current valuedecrement(amount: int = 1)=> subtracts from the current value
Calculation(name=None, function=None)=> creates a new instance of a calculation metric
function=> set the function used the calculation.- The function is to be a parameterless function that returns a numeric value.
- The function can internally reference other items, such as other Statman metrics or access to other resources.
- The function can be a named or lambda function.
read(precision: int = None)=> execute the function, and returns the value rounded based on specified precisionvalue(self)=> execute the function, and returns the value
Rate(name=None, numerator_metric_name=None, denominator_metric_name=None)=> creates a new instance of a rate metric- The
Ratemetric extends theCalculationmetric, where the function is numerator/denominator - The
numerator_metric_nameanddenominator_metric_namerefer to other metrics within the Statman registry
- The
from statman import Statman
Statman.register('expensive-operation-timing',Stopwatch())
stopwatch = Statman.get('expensive-operation-timing')from statman import Statman
Statman.stopwatch('stopwatch-name').start()
# do some expensive operation that you want to measure
Statman.stopwatch('stopwatch-name').read()
print(f'event took {Statman.stopwatch('stopwatch-name').read(precision=1)}s to execute') # event took 1.0s to executefrom statman import Stopwatch
sw = Stopwatch()
sw.start()
# do some expensive operation that you want to measure
delta = sw.stop()
print(f'event took {sw.read(precision=1)}s to execute') # event took 1.0s to executefrom statman import Stopwatch
number_of_events = 1000000
sw = Stopwatch(enable_history=True)
for i in range(0, number_of_events):
sw.start()
# do some expensive operation that you want to measure
sw.stop()
print('number of measurements:', sw.history.count())
print('min:', sw.history.min_value())
print('max:', sw.history.max_value())
print('ave:', sw.history.average_value())
print('mode:', sw.history.mode_value())from statman import Statman
Statman.gauge('number-of-messages-processed')
# in area where something interesting occurs, update gauge
# update can occur using .increment() or .value=
Statman.gauge('number-of-messages_processed').increment()
print('number-of-messages_processed:', Statman.gauge('number-of-messages_processed').value)from statman import Statman
Statman.calculation('messages-per-second').function = lambda: (Statman.gauge('messages-processed').value / Statman.stopwatch('sw').value)
Statman.stopwatch('time-to-process-batch').start()
# code to process batch, incrementing each time message is handles
Statman.gauge('messages-processed').increment()
Statman.stopwatch('sw').stop()
print(Statman.calculation('messages-per-second').value)from statman import Statman
Statman.stopwatch('sw').start()
time.sleep(0.5)
Statman.stopwatch('sw').stop()
Statman.gauge('messages_processed').value = 100
Statman.rate(name='messages_per_second', numerator_metric_name='messages_processed', denominator_metric_name='sw')
print(Statman.rate('messages_per_second').value)